Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
id
string
NUM_NODE
int64
requirements
string
weight
int64
depth
int64
sub_tasks
string
task_category
unknown
finegrained_task_category
unknown
84f6308c-9fcf-4bd4-9c23-135f7dcc12d4
1
The core contributions of the paper When Is Multilinguality a Curse? Language Modeling for 250 High- and Low-Resource Languages have been reproduced.
1,000
0
{"id":"e121dceb-5b9e-492a-8798-1fd52a908cb2","NUM_NODE":2,"requirements":"The reproduction includes pre-training over 1900 monolingual baseline models for 252 languages and estimating language modeling loss as a function of monolingual dataset size, as described in Section 4 (Monolingual Baselines and Metrics).","weigh...
null
null
84f6308c-9fcf-4bd4-9c23-135f7dcc12d4
1
The core contributions of the paper When Is Multilinguality a Curse? Language Modeling for 250 High- and Low-Resource Languages have been reproduced.
1,000
0
{"id":"f00da549-4d65-450a-9066-b18fc3e9411a","NUM_NODE":3,"requirements":"The reproduction includes pre-training over 8400 multilingual language models and evaluates performance variation across languages based on monolingual data size, multilingual data size, linguistic similarity, and model size (up to 45M parameters...
null
null
84f6308c-9fcf-4bd4-9c23-135f7dcc12d4
1
The core contributions of the paper When Is Multilinguality a Curse? Language Modeling for 250 High- and Low-Resource Languages have been reproduced.
1,000
0
{"id":"43194dd7-d73b-4e5a-9c96-c1e086515332","NUM_NODE":4,"requirements":"The reproduction validates the paper\u2019s core findings: (1) moderate multilingual data improves low-resource language modeling (Section 6.1), and (2) high-resource languages consistently suffer in multilingual settings (Section 6.2), with resu...
null
null
84f6308c-9fcf-4bd4-9c23-135f7dcc12d4
1
The core contributions of the paper When Is Multilinguality a Curse? Language Modeling for 250 High- and Low-Resource Languages have been reproduced.
1,000
0
{"id":"dc626e03-a4f5-4b2d-99ae-02b2afc45ab2","NUM_NODE":5,"requirements":"The reproduction confirms that performance effects depend primarily on syntactic similarity of added languages and that vocabulary overlap has only marginal additional impact, as stated in Sections 6.1 and 7.","weight":100,"depth":1,"task_categor...
null
null

AutoPaperBench

Description

AutoPaperBench is a large‑scale benchmark dataset for evaluating the automated replication of AI research papers. It contains 1,986 high‑tier, oral‑level papers published from 2020 to 2025 across 10 top AI conferences (NeurIPS, ICLR, ICML, AAAI, ACL, NAACL, EMNLP, ICCV, CVPR, IJCAI), paired with their official, cleaned code repositories. The dataset supports fully automated, hierarchical evaluation of research reproducibility using agent‑generated rubrics, enabling scalable assessment of LLM and agent capabilities in translating academic papers into executable code and reproducible experiments.

Availability

Full dataset will be uploaded by 2026‑05‑12.

Downloads last month
5