Datasets:

kishormorol
/

researchscope-papers

task stringclasses 5 values	instruction stringclasses 5 values	input stringlengths 105 3.23k	output stringlengths 10 1.93k	paper_id stringlengths 15 150	venue stringclasses 16 values	year int64 1.99k 2.03k	source_type stringclasses 3 values
summarize	Summarize this research paper in 2-3 sentences.	Title: Neuroplastic Expansion in Deep Reinforcement Learning Venue: ICLR 2025 Abstract: The loss of plasticity in learning agents, analogous to the solidification of neural pathways in biological brains, significantly impedes learning and adaptation in reinforcement learning due to its non-stationary nature. To address...	The loss of plasticity in learning agents, analogous to the solidification of neural pathways in biological brains, significantly impedes learning and adaptation in reinforcement learning due to its non-stationary nature. To address this fundamental challenge, we propose a novel approach, Neuroplastic Expansion (NE),...	openreview:20qZK2T7fa	ICLR	2,025	conference
contribution	What is the key contribution of this paper?	Title: Neuroplastic Expansion in Deep Reinforcement Learning Venue: ICLR 2025 Abstract: The loss of plasticity in learning agents, analogous to the solidification of neural pathways in biological brains, significantly impedes learning and adaptation in reinforcement learning due to its non-stationary nature. To address...	To address this fundamental challenge, we propose a novel approach, Neuroplastic Expansion (NE), inspired by cortical expansion in cognitive science.	openreview:20qZK2T7fa	ICLR	2,025	conference
importance	Why does this paper matter to the research community?	Title: Neuroplastic Expansion in Deep Reinforcement Learning Venue: ICLR 2025 Abstract: The loss of plasticity in learning agents, analogous to the solidification of neural pathways in biological brains, significantly impedes learning and adaptation in reinforcement learning due to its non-stationary nature. To address...	Neuroplastic Expansion in Deep Reinforcement Learning contributes to reinforcement learning — a key paradigm for training agents that make sequential decisions.	openreview:20qZK2T7fa	ICLR	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: Neuroplastic Expansion in Deep Reinforcement Learning Venue: ICLR 2025 Abstract: The loss of plasticity in learning agents, analogous to the solidification of neural pathways in biological brains, significantly impedes learning and adaptation in reinforcement learning due to its non-stationary nature. To address...	In plain terms: The loss of plasticity in learning agents, analogous to the solidification of neural pathways in biological brains, significantly impedes learning and adaptation in reinforcement learning due to its non-stationary nature. To address this fundamental challenge, we propose a novel approach, *Neuroplastic ...	openreview:20qZK2T7fa	ICLR	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: Neuroplastic Expansion in Deep Reinforcement Learning Venue: ICLR 2025 Abstract: The loss of plasticity in learning agents, analogous to the solidification of neural pathways in biological brains, significantly impedes learning and adaptation in reinforcement learning due to its non-stationary nature. To address...	To address this fundamental challenge, we propose a novel approach, Neuroplastic Expansion (NE), inspired by cortical…	openreview:20qZK2T7fa	ICLR	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Meta Flow Matching: Integrating Vector Fields on the Wasserstein Manifold Venue: ICLR 2025 Abstract: Numerous biological and physical processes can be modeled as systems of interacting entities evolving continuously over time, e.g. the dynamics of communicating cells or physical particles. Learning the dynamics ...	Numerous biological and physical processes can be modeled as systems of interacting entities evolving continuously over time, e.g. the dynamics of communicating cells or physical particles.	openreview:9SYczU3Qgm	ICLR	2,025	conference
contribution	What is the key contribution of this paper?	Title: Meta Flow Matching: Integrating Vector Fields on the Wasserstein Manifold Venue: ICLR 2025 Abstract: Numerous biological and physical processes can be modeled as systems of interacting entities evolving continuously over time, e.g. the dynamics of communicating cells or physical particles. Learning the dynamics ...	Learning the dynamics of such systems is essential for predicting the temporal evolution of populations across novel samples and unseen environments.	openreview:9SYczU3Qgm	ICLR	2,025	conference
importance	Why does this paper matter to the research community?	Title: Meta Flow Matching: Integrating Vector Fields on the Wasserstein Manifold Venue: ICLR 2025 Abstract: Numerous biological and physical processes can be modeled as systems of interacting entities evolving continuously over time, e.g. the dynamics of communicating cells or physical particles. Learning the dynamics ...	Meta Flow Matching: Integrating Vector Fields on the Wasserstein Manifold broadens the applicability of graph neural networks for structured data and relational reasoning.	openreview:9SYczU3Qgm	ICLR	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: Meta Flow Matching: Integrating Vector Fields on the Wasserstein Manifold Venue: ICLR 2025 Abstract: Numerous biological and physical processes can be modeled as systems of interacting entities evolving continuously over time, e.g. the dynamics of communicating cells or physical particles. Learning the dynamics ...	In plain terms: Numerous biological and physical processes can be modeled as systems of interacting entities evolving continuously over time, e.g. the dynamics of communicating cells or physical particles. Learning the dynamics of such systems is essential for predicting the temporal evolution of populations across nov...	openreview:9SYczU3Qgm	ICLR	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: Meta Flow Matching: Integrating Vector Fields on the Wasserstein Manifold Venue: ICLR 2025 Abstract: Numerous biological and physical processes can be modeled as systems of interacting entities evolving continuously over time, e.g. the dynamics of communicating cells or physical particles. Learning the dynamics ...	Learning the dynamics of such systems is essential for predicting the temporal evolution of populations across novel…	openreview:9SYczU3Qgm	ICLR	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Overcoming Lower-Level Constraints in Bilevel Optimization: A Novel Approach with Regularized Gap Functions Venue: ICLR 2025 Abstract: Constrained bilevel optimization tackles nested structures present in constrained learning tasks like constrained meta-learning, adversarial learning, and distributed bilevel opt...	Constrained bilevel optimization tackles nested structures present in constrained learning tasks like constrained meta-learning, adversarial learning, and distributed bilevel optimization. However, existing bilevel optimization methods mostly are typically restricted to specific constraint settings, such as linear lowe...	openreview:cyPMEXdqQ2	ICLR	2,025	conference
contribution	What is the key contribution of this paper?	Title: Overcoming Lower-Level Constraints in Bilevel Optimization: A Novel Approach with Regularized Gap Functions Venue: ICLR 2025 Abstract: Constrained bilevel optimization tackles nested structures present in constrained learning tasks like constrained meta-learning, adversarial learning, and distributed bilevel opt...	Constrained bilevel optimization tackles nested structures present in constrained learning tasks like constrained meta-learning, adversarial learning, and distributed bilevel optimization.	openreview:cyPMEXdqQ2	ICLR	2,025	conference
importance	Why does this paper matter to the research community?	Title: Overcoming Lower-Level Constraints in Bilevel Optimization: A Novel Approach with Regularized Gap Functions Venue: ICLR 2025 Abstract: Constrained bilevel optimization tackles nested structures present in constrained learning tasks like constrained meta-learning, adversarial learning, and distributed bilevel opt...	Overcoming Lower-Level Constraints in Bilevel Optimization: A Novel Approach with Regularized Gap Functions represents a meaningful contribution to its field and opens new research directions.	openreview:cyPMEXdqQ2	ICLR	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: Overcoming Lower-Level Constraints in Bilevel Optimization: A Novel Approach with Regularized Gap Functions Venue: ICLR 2025 Abstract: Constrained bilevel optimization tackles nested structures present in constrained learning tasks like constrained meta-learning, adversarial learning, and distributed bilevel opt...	In plain terms: Constrained bilevel optimization tackles nested structures present in constrained learning tasks like constrained meta-learning, adversarial learning, and distributed bilevel optimization. However, existing bilevel optimization methods mostly are typically restricted to specific constraint settings, suc...	openreview:cyPMEXdqQ2	ICLR	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: Overcoming Lower-Level Constraints in Bilevel Optimization: A Novel Approach with Regularized Gap Functions Venue: ICLR 2025 Abstract: Constrained bilevel optimization tackles nested structures present in constrained learning tasks like constrained meta-learning, adversarial learning, and distributed bilevel opt...	Constrained bilevel optimization tackles nested structures present in constrained learning tasks like constrained…	openreview:cyPMEXdqQ2	ICLR	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: SVDQuant: Absorbing Outliers by Low-Rank Component for 4-Bit Diffusion Models Venue: ICLR 2025 Abstract: Diffusion models can effectively generate high-quality images. However, as they scale, rising memory demands and higher latency pose substantial deployment challenges. In this work, we aim to accelerate diff...	Diffusion models can effectively generate high-quality images. However, as they scale, rising memory demands and higher latency pose substantial deployment challenges.	openreview:vWR3KuiQur	ICLR	2,025	conference
contribution	What is the key contribution of this paper?	Title: SVDQuant: Absorbing Outliers by Low-Rank Component for 4-Bit Diffusion Models Venue: ICLR 2025 Abstract: Diffusion models can effectively generate high-quality images. However, as they scale, rising memory demands and higher latency pose substantial deployment challenges. In this work, we aim to accelerate diff...	To overcome this limitation, we propose SVDQuant, a new 4-bit quantization paradigm.	openreview:vWR3KuiQur	ICLR	2,025	conference
importance	Why does this paper matter to the research community?	Title: SVDQuant: Absorbing Outliers by Low-Rank Component for 4-Bit Diffusion Models Venue: ICLR 2025 Abstract: Diffusion models can effectively generate high-quality images. However, as they scale, rising memory demands and higher latency pose substantial deployment challenges. In this work, we aim to accelerate diff...	SVDQuant: Absorbing Outliers by Low-Rank Component for 4-Bit Diffusion Models pushes the frontier of generative AI, enabling higher-quality and more controllable content creation.	openreview:vWR3KuiQur	ICLR	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: SVDQuant: Absorbing Outliers by Low-Rank Component for 4-Bit Diffusion Models Venue: ICLR 2025 Abstract: Diffusion models can effectively generate high-quality images. However, as they scale, rising memory demands and higher latency pose substantial deployment challenges. In this work, we aim to accelerate diff...	In plain terms: Diffusion models can effectively generate high-quality images. However, as they scale, rising memory demands and higher latency pose substantial deployment challenges. In this work, we aim to accelerate diffusion models by quantizing their weights and activations to 4 bits.	openreview:vWR3KuiQur	ICLR	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: SVDQuant: Absorbing Outliers by Low-Rank Component for 4-Bit Diffusion Models Venue: ICLR 2025 Abstract: Diffusion models can effectively generate high-quality images. However, as they scale, rising memory demands and higher latency pose substantial deployment challenges. In this work, we aim to accelerate diff...	To overcome this limitation, we propose SVDQuant, a new 4-bit quantization paradigm.	openreview:vWR3KuiQur	ICLR	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: BAMDP Shaping: a Unified Framework for Intrinsic Motivation and Reward Shaping Venue: ICLR 2025 Abstract: Intrinsic motivation and reward shaping guide reinforcement learning (RL) agents by adding pseudo-rewards, which can lead to useful emergent behaviors. However, they can also encourage counterproductive expl...	Intrinsic motivation and reward shaping guide reinforcement learning (RL) agents by adding pseudo-rewards, which can lead to useful emergent behaviors. However, they can also encourage counterproductive exploits, e.g., fixation with noisy TV screens.	openreview:tijmpS9Vy2	ICLR	2,025	conference
contribution	What is the key contribution of this paper?	Title: BAMDP Shaping: a Unified Framework for Intrinsic Motivation and Reward Shaping Venue: ICLR 2025 Abstract: Intrinsic motivation and reward shaping guide reinforcement learning (RL) agents by adding pseudo-rewards, which can lead to useful emergent behaviors. However, they can also encourage counterproductive expl...	We extend potential-based shaping theory to prove BAMDP Potential-based shaping Functions (BAMPFs) are immune to reward-hacking (convergence to behaviors maximizing composite rewards to the detriment of real rewards) in meta-RL, and show empirically how a BAMPF helps a meta-RL agent learn optimal RL algorithms for a Be...	openreview:tijmpS9Vy2	ICLR	2,025	conference
importance	Why does this paper matter to the research community?	Title: BAMDP Shaping: a Unified Framework for Intrinsic Motivation and Reward Shaping Venue: ICLR 2025 Abstract: Intrinsic motivation and reward shaping guide reinforcement learning (RL) agents by adding pseudo-rewards, which can lead to useful emergent behaviors. However, they can also encourage counterproductive expl...	BAMDP Shaping: a Unified Framework for Intrinsic Motivation and Reward Shaping contributes to reinforcement learning — a key paradigm for training agents that make sequential decisions.	openreview:tijmpS9Vy2	ICLR	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: BAMDP Shaping: a Unified Framework for Intrinsic Motivation and Reward Shaping Venue: ICLR 2025 Abstract: Intrinsic motivation and reward shaping guide reinforcement learning (RL) agents by adding pseudo-rewards, which can lead to useful emergent behaviors. However, they can also encourage counterproductive expl...	In plain terms: Intrinsic motivation and reward shaping guide reinforcement learning (RL) agents by adding pseudo-rewards, which can lead to useful emergent behaviors. However, they can also encourage counterproductive exploits, e.g., fixation with noisy TV screens. Here we provide a theoretical model which anticipates...	openreview:tijmpS9Vy2	ICLR	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: BAMDP Shaping: a Unified Framework for Intrinsic Motivation and Reward Shaping Venue: ICLR 2025 Abstract: Intrinsic motivation and reward shaping guide reinforcement learning (RL) agents by adding pseudo-rewards, which can lead to useful emergent behaviors. However, they can also encourage counterproductive expl...	We extend potential-based shaping theory to prove BAMDP Potential-based shaping Functions (BAMPFs) are immune to…	openreview:tijmpS9Vy2	ICLR	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments Venue: COLM 2025 Abstract: The deployment of large-scale models, such as large language models (LLMs), incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalabili...	The deployment of large-scale models, such as large language models (LLMs), incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalability and data security, there is a growing shift towards decentralized systems for model deployment, where choosing ef...	openreview:eLWn2XVMHA	COLM	2,025	conference
contribution	What is the key contribution of this paper?	Title: Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments Venue: COLM 2025 Abstract: The deployment of large-scale models, such as large language models (LLMs), incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalabili...	We demonstrate that our meta-learning framework not only streamlines the decision-making process but also consistently outperforms conventional methods in terms of efficiency and performance.	openreview:eLWn2XVMHA	COLM	2,025	conference
importance	Why does this paper matter to the research community?	Title: Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments Venue: COLM 2025 Abstract: The deployment of large-scale models, such as large language models (LLMs), incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalabili...	Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments represents a meaningful contribution to its field and opens new research directions.	openreview:eLWn2XVMHA	COLM	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments Venue: COLM 2025 Abstract: The deployment of large-scale models, such as large language models (LLMs), incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalabili...	In plain terms: The deployment of large-scale models, such as large language models (LLMs), incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalability and data security, there is a growing shift towards decentralized systems for model deployment, w...	openreview:eLWn2XVMHA	COLM	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments Venue: COLM 2025 Abstract: The deployment of large-scale models, such as large language models (LLMs), incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalabili...	We demonstrate that our meta-learning framework not only streamlines the decision-making process but also consistently…	openreview:eLWn2XVMHA	COLM	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Efficient Streaming Language Models with Attention Sinks Venue: ICLR 2024 Abstract: Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching p...	Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching previous tokens' Key and Value states (KV) consumes extensive memory.	openreview:NG7sS51zVF	ICLR	2,024	conference
contribution	What is the key contribution of this paper?	Title: Efficient Streaming Language Models with Attention Sinks Venue: ICLR 2024 Abstract: Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching p...	Window attention, where only the most recent KVs are cached, is a natural approach --- but we show that it fails when the text length surpasses the cache size.	openreview:NG7sS51zVF	ICLR	2,024	conference
importance	Why does this paper matter to the research community?	Title: Efficient Streaming Language Models with Attention Sinks Venue: ICLR 2024 Abstract: Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching p...	Efficient Streaming Language Models with Attention Sinks represents a meaningful contribution to its field and opens new research directions.	openreview:NG7sS51zVF	ICLR	2,024	conference
plain	Explain this paper in plain English for a non-expert.	Title: Efficient Streaming Language Models with Attention Sinks Venue: ICLR 2024 Abstract: Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching p...	In plain terms: Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching previous tokens' Key and Value states (KV) consumes extensive memory. Second...	openreview:NG7sS51zVF	ICLR	2,024	conference
takeaway	Give a one-line takeaway from this paper.	Title: Efficient Streaming Language Models with Attention Sinks Venue: ICLR 2024 Abstract: Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching p...	Window attention, where only the most recent KVs are cached, is a natural approach --- but we show that it fails when…	openreview:NG7sS51zVF	ICLR	2,024	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Topology-Aware Vision Transformers for Enhanced Scene Recognition Venue: AAAI 2026 Abstract: Scene recognition (SR) is a fundamental task in computer vision (CV). In recent years, Transformer-based methods have achieved remarkable success in scene recognition tasks. Most existing approaches primarily rely on vis...	Scene recognition (SR) is a fundamental task in computer vision (CV). In recent years, Transformer-based methods have achieved remarkable success in scene recognition tasks.	s2:b1ca730de7c3282859764d71f96575fc4388aa1c	AAAI	2,026	conference
contribution	What is the key contribution of this paper?	Title: Topology-Aware Vision Transformers for Enhanced Scene Recognition Venue: AAAI 2026 Abstract: Scene recognition (SR) is a fundamental task in computer vision (CV). In recent years, Transformer-based methods have achieved remarkable success in scene recognition tasks. Most existing approaches primarily rely on vis...	To this end, we propose Topology Attention Network for Scene Recognition (TANSR), an innovative method that leverages topological relationships from graphs to guide scene recognition.	s2:b1ca730de7c3282859764d71f96575fc4388aa1c	AAAI	2,026	conference
importance	Why does this paper matter to the research community?	Title: Topology-Aware Vision Transformers for Enhanced Scene Recognition Venue: AAAI 2026 Abstract: Scene recognition (SR) is a fundamental task in computer vision (CV). In recent years, Transformer-based methods have achieved remarkable success in scene recognition tasks. Most existing approaches primarily rely on vis...	Topology-Aware Vision Transformers for Enhanced Scene Recognition extends Transformer capabilities — the architecture underpinning virtually every modern NLP and vision model.	s2:b1ca730de7c3282859764d71f96575fc4388aa1c	AAAI	2,026	conference
plain	Explain this paper in plain English for a non-expert.	Title: Topology-Aware Vision Transformers for Enhanced Scene Recognition Venue: AAAI 2026 Abstract: Scene recognition (SR) is a fundamental task in computer vision (CV). In recent years, Transformer-based methods have achieved remarkable success in scene recognition tasks. Most existing approaches primarily rely on vis...	In plain terms: Scene recognition (SR) is a fundamental task in computer vision (CV). In recent years, Transformer-based methods have achieved remarkable success in scene recognition tasks. Most existing approaches primarily rely on visual features, while failing to effectively model the structural relationships within...	s2:b1ca730de7c3282859764d71f96575fc4388aa1c	AAAI	2,026	conference
takeaway	Give a one-line takeaway from this paper.	Title: Topology-Aware Vision Transformers for Enhanced Scene Recognition Venue: AAAI 2026 Abstract: Scene recognition (SR) is a fundamental task in computer vision (CV). In recent years, Transformer-based methods have achieved remarkable success in scene recognition tasks. Most existing approaches primarily rely on vis...	To this end, we propose Topology Attention Network for Scene Recognition (TANSR), an innovative method that leverages…	s2:b1ca730de7c3282859764d71f96575fc4388aa1c	AAAI	2,026	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Efficient Heterogeneous Meta-Learning via Channel Shuffling Modulation Venue: ICLR 2024 Abstract: We tackle the problem of meta-learning across heterogenous tasks. This problem seeks to extract and generalize transferable meta-knowledge through streaming task sets from a multi-modal task distribution. The extrac...	We tackle the problem of meta-learning across heterogenous tasks. This problem seeks to extract and generalize transferable meta-knowledge through streaming task sets from a multi-modal task distribution.	openreview:QiJuMJl0QS	ICLR	2,024	conference
contribution	What is the key contribution of this paper?	Title: Efficient Heterogeneous Meta-Learning via Channel Shuffling Modulation Venue: ICLR 2024 Abstract: We tackle the problem of meta-learning across heterogenous tasks. This problem seeks to extract and generalize transferable meta-knowledge through streaming task sets from a multi-modal task distribution. The extrac...	To overcome this weakness, we propose a new heterogeneous meta-learning strategy that efficiently captures the multi-modality of the task distribution via modulating the routing between convolution channels in the network, instead of directly modulating the network weights.	openreview:QiJuMJl0QS	ICLR	2,024	conference
importance	Why does this paper matter to the research community?	Title: Efficient Heterogeneous Meta-Learning via Channel Shuffling Modulation Venue: ICLR 2024 Abstract: We tackle the problem of meta-learning across heterogenous tasks. This problem seeks to extract and generalize transferable meta-knowledge through streaming task sets from a multi-modal task distribution. The extrac...	Efficient Heterogeneous Meta-Learning via Channel Shuffling Modulation bridges vision and language, enabling richer AI applications that reason across modalities.	openreview:QiJuMJl0QS	ICLR	2,024	conference
plain	Explain this paper in plain English for a non-expert.	Title: Efficient Heterogeneous Meta-Learning via Channel Shuffling Modulation Venue: ICLR 2024 Abstract: We tackle the problem of meta-learning across heterogenous tasks. This problem seeks to extract and generalize transferable meta-knowledge through streaming task sets from a multi-modal task distribution. The extrac...	In plain terms: We tackle the problem of meta-learning across heterogenous tasks. This problem seeks to extract and generalize transferable meta-knowledge through streaming task sets from a multi-modal task distribution. The extracted meta-knowledge can be used to create predictors for new tasks using a small number of...	openreview:QiJuMJl0QS	ICLR	2,024	conference
takeaway	Give a one-line takeaway from this paper.	Title: Efficient Heterogeneous Meta-Learning via Channel Shuffling Modulation Venue: ICLR 2024 Abstract: We tackle the problem of meta-learning across heterogenous tasks. This problem seeks to extract and generalize transferable meta-knowledge through streaming task sets from a multi-modal task distribution. The extrac...	To overcome this weakness, we propose a new heterogeneous meta-learning strategy that efficiently captures the…	openreview:QiJuMJl0QS	ICLR	2,024	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: LeakAgent: RL-based Red-teaming Agent for LLM Privacy Leakage Venue: COLM 2025 Abstract: Recent studies have discovered that large language models (LLM) may be ``fooled'' to output private information, including training data, system prompts, and personally identifiable information, under carefully crafted adver...	Recent studies have discovered that large language models (LLM) may be ``fooled'' to output private information, including training data, system prompts, and personally identifiable information, under carefully crafted adversarial prompts. Existing red-teaming approaches for privacy leakage either rely on manual effort...	openreview:WIfns41MAb	COLM	2,025	conference
contribution	What is the key contribution of this paper?	Title: LeakAgent: RL-based Red-teaming Agent for LLM Privacy Leakage Venue: COLM 2025 Abstract: Recent studies have discovered that large language models (LLM) may be ``fooled'' to output private information, including training data, system prompts, and personally identifiable information, under carefully crafted adver...	We propose LeakAgent, a novel black-box red-teaming framework for LLM privacy leakage.	openreview:WIfns41MAb	COLM	2,025	conference
importance	Why does this paper matter to the research community?	Title: LeakAgent: RL-based Red-teaming Agent for LLM Privacy Leakage Venue: COLM 2025 Abstract: Recent studies have discovered that large language models (LLM) may be ``fooled'' to output private information, including training data, system prompts, and personally identifiable information, under carefully crafted adver...	LeakAgent: RL-based Red-teaming Agent for LLM Privacy Leakage represents a meaningful contribution to its field and opens new research directions.	openreview:WIfns41MAb	COLM	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: LeakAgent: RL-based Red-teaming Agent for LLM Privacy Leakage Venue: COLM 2025 Abstract: Recent studies have discovered that large language models (LLM) may be ``fooled'' to output private information, including training data, system prompts, and personally identifiable information, under carefully crafted adver...	In plain terms: Recent studies have discovered that large language models (LLM) may be ``fooled'' to output private information, including training data, system prompts, and personally identifiable information, under carefully crafted adversarial prompts. Existing red-teaming approaches for privacy leakage either rely ...	openreview:WIfns41MAb	COLM	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: LeakAgent: RL-based Red-teaming Agent for LLM Privacy Leakage Venue: COLM 2025 Abstract: Recent studies have discovered that large language models (LLM) may be ``fooled'' to output private information, including training data, system prompts, and personally identifiable information, under carefully crafted adver...	We propose LeakAgent, a novel black-box red-teaming framework for LLM privacy leakage.	openreview:WIfns41MAb	COLM	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Zeroth-Order Optimization Meets Human Feedback: Provable Learning via Ranking Oracles Venue: ICLR 2024 Abstract: In this study, we delve into an emerging optimization challenge involving a black-box objective function that can only be gauged via a ranking oracle—a situation frequently encountered in real-world s...	In this study, we delve into an emerging optimization challenge involving a black-box objective function that can only be gauged via a ranking oracle—a situation frequently encountered in real-world scenarios, especially when the function is evaluated by human judges. A prominent instance of such a situation is Reinfor...	openreview:TVDUVpgu9s	ICLR	2,024	conference
contribution	What is the key contribution of this paper?	Title: Zeroth-Order Optimization Meets Human Feedback: Provable Learning via Ranking Oracles Venue: ICLR 2024 Abstract: In this study, we delve into an emerging optimization challenge involving a black-box objective function that can only be gauged via a ranking oracle—a situation frequently encountered in real-world s...	We introduce ZO-RankSGD, an innovative zeroth-order optimization algorithm designed to tackle this optimization problem, accompanied by theoretical assurances.	openreview:TVDUVpgu9s	ICLR	2,024	conference
importance	Why does this paper matter to the research community?	Title: Zeroth-Order Optimization Meets Human Feedback: Provable Learning via Ranking Oracles Venue: ICLR 2024 Abstract: In this study, we delve into an emerging optimization challenge involving a black-box objective function that can only be gauged via a ranking oracle—a situation frequently encountered in real-world s...	Zeroth-Order Optimization Meets Human Feedback: Provable Learning via Ranking Oracles contributes to reinforcement learning — a key paradigm for training agents that make sequential decisions.	openreview:TVDUVpgu9s	ICLR	2,024	conference
plain	Explain this paper in plain English for a non-expert.	Title: Zeroth-Order Optimization Meets Human Feedback: Provable Learning via Ranking Oracles Venue: ICLR 2024 Abstract: In this study, we delve into an emerging optimization challenge involving a black-box objective function that can only be gauged via a ranking oracle—a situation frequently encountered in real-world s...	In plain terms: In this study, we delve into an emerging optimization challenge involving a black-box objective function that can only be gauged via a ranking oracle—a situation frequently encountered in real-world scenarios, especially when the function is evaluated by human judges. A prominent instance of such a situ...	openreview:TVDUVpgu9s	ICLR	2,024	conference
takeaway	Give a one-line takeaway from this paper.	Title: Zeroth-Order Optimization Meets Human Feedback: Provable Learning via Ranking Oracles Venue: ICLR 2024 Abstract: In this study, we delve into an emerging optimization challenge involving a black-box objective function that can only be gauged via a ranking oracle—a situation frequently encountered in real-world s...	We introduce ZO-RankSGD, an innovative zeroth-order optimization algorithm designed to tackle this optimization…	openreview:TVDUVpgu9s	ICLR	2,024	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: SRL: Scaling Distributed Reinforcement Learning to Over Ten Thousand Cores Venue: ICLR 2024 Abstract: The ever-growing complexity of reinforcement learning (RL) tasks demands a distributed system to efficiently generate and process a massive amount of data. However, existing open-source libraries suffer from var...	The ever-growing complexity of reinforcement learning (RL) tasks demands a distributed system to efficiently generate and process a massive amount of data. However, existing open-source libraries suffer from various limitations, which impede their practical use in challenging scenarios where large-scale training is nec...	openreview:lajn1iROCu	ICLR	2,024	conference
contribution	What is the key contribution of this paper?	Title: SRL: Scaling Distributed Reinforcement Learning to Over Ten Thousand Cores Venue: ICLR 2024 Abstract: The ever-growing complexity of reinforcement learning (RL) tasks demands a distributed system to efficiently generate and process a massive amount of data. However, existing open-source libraries suffer from var...	In this paper, we present a novel abstraction on the dataflows of RL training, which unifies diverse RL training applications into a general framework.	openreview:lajn1iROCu	ICLR	2,024	conference
importance	Why does this paper matter to the research community?	Title: SRL: Scaling Distributed Reinforcement Learning to Over Ten Thousand Cores Venue: ICLR 2024 Abstract: The ever-growing complexity of reinforcement learning (RL) tasks demands a distributed system to efficiently generate and process a massive amount of data. However, existing open-source libraries suffer from var...	SRL: Scaling Distributed Reinforcement Learning to Over Ten Thousand Cores contributes to reinforcement learning — a key paradigm for training agents that make sequential decisions.	openreview:lajn1iROCu	ICLR	2,024	conference
plain	Explain this paper in plain English for a non-expert.	Title: SRL: Scaling Distributed Reinforcement Learning to Over Ten Thousand Cores Venue: ICLR 2024 Abstract: The ever-growing complexity of reinforcement learning (RL) tasks demands a distributed system to efficiently generate and process a massive amount of data. However, existing open-source libraries suffer from var...	In plain terms: The ever-growing complexity of reinforcement learning (RL) tasks demands a distributed system to efficiently generate and process a massive amount of data. However, existing open-source libraries suffer from various limitations, which impede their practical use in challenging scenarios where large-scale...	openreview:lajn1iROCu	ICLR	2,024	conference
takeaway	Give a one-line takeaway from this paper.	Title: SRL: Scaling Distributed Reinforcement Learning to Over Ten Thousand Cores Venue: ICLR 2024 Abstract: The ever-growing complexity of reinforcement learning (RL) tasks demands a distributed system to efficiently generate and process a massive amount of data. However, existing open-source libraries suffer from var...	In this paper, we present a novel abstraction on the dataflows of RL training, which unifies diverse RL training…	openreview:lajn1iROCu	ICLR	2,024	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery Venue: ICLR 2025 Abstract: Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamic...	Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamical systems guarantees convergence to a desired state, it often overlooks transient behavior.	openreview:lILEtkWOXD	ICLR	2,025	conference
contribution	What is the key contribution of this paper?	Title: Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery Venue: ICLR 2025 Abstract: Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamic...	We propose a framework for learning policies modeled by contractive dynamical systems, ensuring that all policy rollouts converge regardless of perturbations, and in turn, enable efficient OOS recovery.	openreview:lILEtkWOXD	ICLR	2,025	conference
importance	Why does this paper matter to the research community?	Title: Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery Venue: ICLR 2025 Abstract: Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamic...	Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery represents a meaningful contribution to its field and opens new research directions.	openreview:lILEtkWOXD	ICLR	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery Venue: ICLR 2025 Abstract: Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamic...	In plain terms: Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamical systems guarantees convergence to a desired state, it often overlooks transient behavior. We ...	openreview:lILEtkWOXD	ICLR	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery Venue: ICLR 2025 Abstract: Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamic...	We propose a framework for learning policies modeled by contractive dynamical systems, ensuring that all policy…	openreview:lILEtkWOXD	ICLR	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Meta Inverse Constrained Reinforcement Learning: Convergence Guarantee and Generalization Analysis Venue: ICLR 2024 Abstract: This paper considers the problem of learning the reward function and constraints of an expert from few demonstrations. This problem can be considered as a meta-learning problem where we f...	This paper considers the problem of learning the reward function and constraints of an expert from few demonstrations. This problem can be considered as a meta-learning problem where we first learn meta-priors over reward functions and constraints from other distinct but related tasks and then adapt the learned meta-pr...	openreview:bJ3gFiwRgi	ICLR	2,024	conference
contribution	What is the key contribution of this paper?	Title: Meta Inverse Constrained Reinforcement Learning: Convergence Guarantee and Generalization Analysis Venue: ICLR 2024 Abstract: This paper considers the problem of learning the reward function and constraints of an expert from few demonstrations. This problem can be considered as a meta-learning problem where we f...	We propose a novel algorithm to solve this problem and formally guarantee that the algorithm reaches the set of $\epsilon$-stationary points at the iteration complexity $O(\frac{1}{\epsilon^2})$.	openreview:bJ3gFiwRgi	ICLR	2,024	conference
importance	Why does this paper matter to the research community?	Title: Meta Inverse Constrained Reinforcement Learning: Convergence Guarantee and Generalization Analysis Venue: ICLR 2024 Abstract: This paper considers the problem of learning the reward function and constraints of an expert from few demonstrations. This problem can be considered as a meta-learning problem where we f...	Meta Inverse Constrained Reinforcement Learning: Convergence Guarantee and Generalization Analysis contributes to reinforcement learning — a key paradigm for training agents that make sequential decisions.	openreview:bJ3gFiwRgi	ICLR	2,024	conference
plain	Explain this paper in plain English for a non-expert.	Title: Meta Inverse Constrained Reinforcement Learning: Convergence Guarantee and Generalization Analysis Venue: ICLR 2024 Abstract: This paper considers the problem of learning the reward function and constraints of an expert from few demonstrations. This problem can be considered as a meta-learning problem where we f...	In plain terms: This paper considers the problem of learning the reward function and constraints of an expert from few demonstrations. This problem can be considered as a meta-learning problem where we first learn meta-priors over reward functions and constraints from other distinct but related tasks and then adapt the...	openreview:bJ3gFiwRgi	ICLR	2,024	conference
takeaway	Give a one-line takeaway from this paper.	Title: Meta Inverse Constrained Reinforcement Learning: Convergence Guarantee and Generalization Analysis Venue: ICLR 2024 Abstract: This paper considers the problem of learning the reward function and constraints of an expert from few demonstrations. This problem can be considered as a meta-learning problem where we f...	We propose a novel algorithm to solve this problem and formally guarantee that the algorithm reaches the set of…	openreview:bJ3gFiwRgi	ICLR	2,024	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: qNBO: quasi-Newton Meets Bilevel Optimization Venue: ICLR 2025 Abstract: Bilevel optimization, which addresses challenges in hierarchical learning tasks, has gained significant interest in machine learning. Implementing gradient descent for bilevel optimization presents computational hurdles, notably the need to...	Bilevel optimization, which addresses challenges in hierarchical learning tasks, has gained significant interest in machine learning. Implementing gradient descent for bilevel optimization presents computational hurdles, notably the need to compute the exact lower-level solution and the inverse Hessian of the lower-lev...	openreview:BTOdzCzSRg	ICLR	2,025	conference
contribution	What is the key contribution of this paper?	Title: qNBO: quasi-Newton Meets Bilevel Optimization Venue: ICLR 2025 Abstract: Bilevel optimization, which addresses challenges in hierarchical learning tasks, has gained significant interest in machine learning. Implementing gradient descent for bilevel optimization presents computational hurdles, notably the need to...	In this paper, we introduce a general framework to tackle these computational challenges in a coordinated manner.	openreview:BTOdzCzSRg	ICLR	2,025	conference
importance	Why does this paper matter to the research community?	Title: qNBO: quasi-Newton Meets Bilevel Optimization Venue: ICLR 2025 Abstract: Bilevel optimization, which addresses challenges in hierarchical learning tasks, has gained significant interest in machine learning. Implementing gradient descent for bilevel optimization presents computational hurdles, notably the need to...	qNBO: quasi-Newton Meets Bilevel Optimization unlocks new capabilities from existing models without any additional training.	openreview:BTOdzCzSRg	ICLR	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: qNBO: quasi-Newton Meets Bilevel Optimization Venue: ICLR 2025 Abstract: Bilevel optimization, which addresses challenges in hierarchical learning tasks, has gained significant interest in machine learning. Implementing gradient descent for bilevel optimization presents computational hurdles, notably the need to...	In plain terms: Bilevel optimization, which addresses challenges in hierarchical learning tasks, has gained significant interest in machine learning. Implementing gradient descent for bilevel optimization presents computational hurdles, notably the need to compute the exact lower-level solution and the inverse Hessian ...	openreview:BTOdzCzSRg	ICLR	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: qNBO: quasi-Newton Meets Bilevel Optimization Venue: ICLR 2025 Abstract: Bilevel optimization, which addresses challenges in hierarchical learning tasks, has gained significant interest in machine learning. Implementing gradient descent for bilevel optimization presents computational hurdles, notably the need to...	In this paper, we introduce a general framework to tackle these computational challenges in a coordinated manner.	openreview:BTOdzCzSRg	ICLR	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Cheaply Estimating Inference Efficiency Metrics for Autoregressive Transformer Models Venue: NeurIPS 2023 Abstract: Large language models (LLMs) are highly capable but also computationally expensive. Characterizing the _fundamental tradeoff_ between inference efficiency and model capabilities is thus important,...	Large language models (LLMs) are highly capable but also computationally expensive. Characterizing the _fundamental tradeoff_ between inference efficiency and model capabilities is thus important, but requires an efficiency metric that is comparable across models from different providers.	openreview:RJpAz15D0S	NeurIPS	2,023	conference
contribution	What is the key contribution of this paper?	Title: Cheaply Estimating Inference Efficiency Metrics for Autoregressive Transformer Models Venue: NeurIPS 2023 Abstract: Large language models (LLMs) are highly capable but also computationally expensive. Characterizing the _fundamental tradeoff_ between inference efficiency and model capabilities is thus important,...	We propose a new metric for inference efficiency called _idealized runtime_, that puts models on equal footing as though they were served on uniform hardware and software without performance contention, and a cost model to efficiently estimate this metric for autoregressive Transformer models.	openreview:RJpAz15D0S	NeurIPS	2,023	conference
importance	Why does this paper matter to the research community?	Title: Cheaply Estimating Inference Efficiency Metrics for Autoregressive Transformer Models Venue: NeurIPS 2023 Abstract: Large language models (LLMs) are highly capable but also computationally expensive. Characterizing the _fundamental tradeoff_ between inference efficiency and model capabilities is thus important,...	Cheaply Estimating Inference Efficiency Metrics for Autoregressive Transformer Models extends Transformer capabilities — the architecture underpinning virtually every modern NLP and vision model.	openreview:RJpAz15D0S	NeurIPS	2,023	conference
plain	Explain this paper in plain English for a non-expert.	Title: Cheaply Estimating Inference Efficiency Metrics for Autoregressive Transformer Models Venue: NeurIPS 2023 Abstract: Large language models (LLMs) are highly capable but also computationally expensive. Characterizing the _fundamental tradeoff_ between inference efficiency and model capabilities is thus important,...	In plain terms: Large language models (LLMs) are highly capable but also computationally expensive. Characterizing the _fundamental tradeoff_ between inference efficiency and model capabilities is thus important, but requires an efficiency metric that is comparable across models from different providers. Unfortunately,...	openreview:RJpAz15D0S	NeurIPS	2,023	conference
takeaway	Give a one-line takeaway from this paper.	Title: Cheaply Estimating Inference Efficiency Metrics for Autoregressive Transformer Models Venue: NeurIPS 2023 Abstract: Large language models (LLMs) are highly capable but also computationally expensive. Characterizing the _fundamental tradeoff_ between inference efficiency and model capabilities is thus important,...	We propose a new metric for inference efficiency called _idealized runtime_, that puts models on equal footing as…	openreview:RJpAz15D0S	NeurIPS	2,023	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Meta-rater: A Multi-dimensional Data Selection Method for Pre-training Language Models Venue: ACL 2025 Abstract: The composition of pre-training datasets for large language models (LLMs) remains largely undisclosed, hindering transparency and efforts to optimize data quality---a critical driver of model performa...	The composition of pre-training datasets for large language models (LLMs) remains largely undisclosed, hindering transparency and efforts to optimize data quality---a critical driver of model performance. Current data selection methods, such as natural language quality assessments, diversity-based filters, and classifi...	acl:2025.acl-long.533	ACL	2,025	conference
contribution	What is the key contribution of this paper?	Title: Meta-rater: A Multi-dimensional Data Selection Method for Pre-training Language Models Venue: ACL 2025 Abstract: The composition of pre-training datasets for large language models (LLMs) remains largely undisclosed, hindering transparency and efforts to optimize data quality---a critical driver of model performa...	To address these gaps, we propose four dimensions to evaluate data quality: professionalism, readability, reasoning, and cleanliness.	acl:2025.acl-long.533	ACL	2,025	conference
importance	Why does this paper matter to the research community?	Title: Meta-rater: A Multi-dimensional Data Selection Method for Pre-training Language Models Venue: ACL 2025 Abstract: The composition of pre-training datasets for large language models (LLMs) remains largely undisclosed, hindering transparency and efforts to optimize data quality---a critical driver of model performa...	Meta-rater: A Multi-dimensional Data Selection Method for Pre-training Language Models represents a meaningful contribution to its field and opens new research directions.	acl:2025.acl-long.533	ACL	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: Meta-rater: A Multi-dimensional Data Selection Method for Pre-training Language Models Venue: ACL 2025 Abstract: The composition of pre-training datasets for large language models (LLMs) remains largely undisclosed, hindering transparency and efforts to optimize data quality---a critical driver of model performa...	In plain terms: The composition of pre-training datasets for large language models (LLMs) remains largely undisclosed, hindering transparency and efforts to optimize data quality---a critical driver of model performance. Current data selection methods, such as natural language quality assessments, diversity-based filte...	acl:2025.acl-long.533	ACL	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: Meta-rater: A Multi-dimensional Data Selection Method for Pre-training Language Models Venue: ACL 2025 Abstract: The composition of pre-training datasets for large language models (LLMs) remains largely undisclosed, hindering transparency and efforts to optimize data quality---a critical driver of model performa...	To address these gaps, we propose four dimensions to evaluate data quality: professionalism, readability, reasoning,…	acl:2025.acl-long.533	ACL	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Faster Accelerated First-order Methods for Convex Optimization with Strongly Convex Function Constraints Venue: NeurIPS 2024 Abstract: In this paper, we introduce faster accelerated primal-dual algorithms for minimizing a convex function subject to strongly convex function constraints. Prior to our work, the be...	In this paper, we introduce faster accelerated primal-dual algorithms for minimizing a convex function subject to strongly convex function constraints. Prior to our work, the best complexity bound was $\mathcal{O}(1/{\varepsilon})$, regardless of the strong convexity of the constraint function.	openreview:pG380vLYRU	NeurIPS	2,024	conference
contribution	What is the key contribution of this paper?	Title: Faster Accelerated First-order Methods for Convex Optimization with Strongly Convex Function Constraints Venue: NeurIPS 2024 Abstract: In this paper, we introduce faster accelerated primal-dual algorithms for minimizing a convex function subject to strongly convex function constraints. Prior to our work, the be...	In this paper, we introduce faster accelerated primal-dual algorithms for minimizing a convex function subject to strongly convex function constraints.	openreview:pG380vLYRU	NeurIPS	2,024	conference
importance	Why does this paper matter to the research community?	Title: Faster Accelerated First-order Methods for Convex Optimization with Strongly Convex Function Constraints Venue: NeurIPS 2024 Abstract: In this paper, we introduce faster accelerated primal-dual algorithms for minimizing a convex function subject to strongly convex function constraints. Prior to our work, the be...	Faster Accelerated First-order Methods for Convex Optimization with Strongly Convex Function Constraints represents a meaningful contribution to its field and opens new research directions.	openreview:pG380vLYRU	NeurIPS	2,024	conference
plain	Explain this paper in plain English for a non-expert.	Title: Faster Accelerated First-order Methods for Convex Optimization with Strongly Convex Function Constraints Venue: NeurIPS 2024 Abstract: In this paper, we introduce faster accelerated primal-dual algorithms for minimizing a convex function subject to strongly convex function constraints. Prior to our work, the be...	In plain terms: In this paper, we introduce faster accelerated primal-dual algorithms for minimizing a convex function subject to strongly convex function constraints. Prior to our work, the best complexity bound was $\mathcal{O}(1/{\varepsilon})$, regardless of the strong convexity of the constraint function. It is un...	openreview:pG380vLYRU	NeurIPS	2,024	conference
takeaway	Give a one-line takeaway from this paper.	Title: Faster Accelerated First-order Methods for Convex Optimization with Strongly Convex Function Constraints Venue: NeurIPS 2024 Abstract: In this paper, we introduce faster accelerated primal-dual algorithms for minimizing a convex function subject to strongly convex function constraints. Prior to our work, the be...	In this paper, we introduce faster accelerated primal-dual algorithms for minimizing a convex function subject to…	openreview:pG380vLYRU	NeurIPS	2,024	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: Cross-Lingual Auto Evaluation for Assessing Multilingual LLMs Venue: ACL 2025 Abstract: Evaluating machine-generated text remains a significant challenge in NLP, especially for non-English languages. Current methodologies, including automated metrics, human assessments, and LLM-based evaluations, predominantly f...	Evaluating machine-generated text remains a significant challenge in NLP, especially for non-English languages. Current methodologies, including automated metrics, human assessments, and LLM-based evaluations, predominantly focus on English, revealing a significant gap in multilingual evaluation frameworks.	acl:2025.acl-long.1419	ACL	2,025	conference
contribution	What is the key contribution of this paper?	Title: Cross-Lingual Auto Evaluation for Assessing Multilingual LLMs Venue: ACL 2025 Abstract: Evaluating machine-generated text remains a significant challenge in NLP, especially for non-English languages. Current methodologies, including automated metrics, human assessments, and LLM-based evaluations, predominantly f...	We introduce the Cross Lingual Auto Evaluation (CIA) Suite, an extensible framework that includes evaluator LLMs (Hercule) and a novel test set (Recon) specifically designed for multilingual evaluation.	acl:2025.acl-long.1419	ACL	2,025	conference
importance	Why does this paper matter to the research community?	Title: Cross-Lingual Auto Evaluation for Assessing Multilingual LLMs Venue: ACL 2025 Abstract: Evaluating machine-generated text remains a significant challenge in NLP, especially for non-English languages. Current methodologies, including automated metrics, human assessments, and LLM-based evaluations, predominantly f...	Cross-Lingual Auto Evaluation for Assessing Multilingual LLMs advances our understanding of large language models, which are reshaping how we build AI systems and interact with knowledge.	acl:2025.acl-long.1419	ACL	2,025	conference
plain	Explain this paper in plain English for a non-expert.	Title: Cross-Lingual Auto Evaluation for Assessing Multilingual LLMs Venue: ACL 2025 Abstract: Evaluating machine-generated text remains a significant challenge in NLP, especially for non-English languages. Current methodologies, including automated metrics, human assessments, and LLM-based evaluations, predominantly f...	In plain terms: Evaluating machine-generated text remains a significant challenge in NLP, especially for non-English languages. Current methodologies, including automated metrics, human assessments, and LLM-based evaluations, predominantly focus on English, revealing a significant gap in multilingual evaluation framewo...	acl:2025.acl-long.1419	ACL	2,025	conference
takeaway	Give a one-line takeaway from this paper.	Title: Cross-Lingual Auto Evaluation for Assessing Multilingual LLMs Venue: ACL 2025 Abstract: Evaluating machine-generated text remains a significant challenge in NLP, especially for non-English languages. Current methodologies, including automated metrics, human assessments, and LLM-based evaluations, predominantly f...	We introduce the Cross Lingual Auto Evaluation (CIA) Suite, an extensible framework that includes evaluator LLMs…	acl:2025.acl-long.1419	ACL	2,025	conference
summarize	Summarize this research paper in 2-3 sentences.	Title: MixLLM: Dynamic Routing in Mixed Large Language Models Venue: NAACL 2025 Abstract: Large Language Models (LLMs) exhibit potential artificial generic intelligence recently, however, their usage is costly with high response latency. Given mixed LLMs with their own strengths and weaknesses, LLM routing aims to iden...	Large Language Models (LLMs) exhibit potential artificial generic intelligence recently, however, their usage is costly with high response latency. Given mixed LLMs with their own strengths and weaknesses, LLM routing aims to identify the most suitable model for each query in the stream to maximize response quality and...	acl:2025.naacl-long.545	NAACL	2,025	conference
contribution	What is the key contribution of this paper?	Title: MixLLM: Dynamic Routing in Mixed Large Language Models Venue: NAACL 2025 Abstract: Large Language Models (LLMs) exhibit potential artificial generic intelligence recently, however, their usage is costly with high response latency. Given mixed LLMs with their own strengths and weaknesses, LLM routing aims to iden...	Our extensive experiments show that MixLLM achieves the best trade-offs in response quality, cost, and latency (97.25\% of GPT-4's quality at 24.18\% of the cost under the time constraint).	acl:2025.naacl-long.545	NAACL	2,025	conference

End of preview. Expand in Data Studio

ResearchScope Papers

Open CS research paper dataset maintained by ResearchScope.

Updated automatically via GitHub Actions.

Stats

30,467 papers (raw metadata)
152,320 instruction-tuning rows
Sources: arXiv, OpenAlex, ACL Anthology, OpenReview, PMLR, CVF, Semantic Scholar
Venues: NeurIPS, ICML, ICLR, ACL, EMNLP, CVPR, AAAI, IJCAI, JMLR, TMLR, TACL, TPAMI, NMI and more

Files

File	Description
`data/papers.jsonl`	Raw paper metadata — title, abstract, authors, venue, year, tags, scores
`data/instruct.jsonl`	Instruction-tuning pairs — summarize, key contribution, why it matters, plain English

Usage

from datasets import load_dataset

# Raw papers
papers = load_dataset("kishormorol/researchscope-papers", "papers", split="train")

# Instruction tuning
instruct = load_dataset("kishormorol/researchscope-papers", "instruct", split="train")

License

Paper metadata is aggregated from open sources. Text content follows the original licenses of each source (arXiv CC0, ACL CC BY, etc.). Dataset schema: CC BY 4.0.

Downloads last month: 20