Rajdeep Borgohain's picture

Rajdeep Borgohain

rbgo

·

RajdeepBorgohain

AI & ML interests

Solving language barriers.

Recent Activity

Reacted to m-ric's post with 👍 2 days ago

🔍 Meta teams use a fine-tuned Llama model to fix production issues in seconds One of Meta's engineering teams shared how they use a fine-tuned small Llama (Llama-2-7B, so not even a very recent model) to identify the root cause of production issues with 42% accuracy. 🤔 42%, is that not too low? ➡️ Usually, whenever there's an issue in production, engineers dive into recent code changes to find the offending commit. At Meta's scale (thousands of daily changes), this is like finding a needle in a haystack. 💡 So when the LLM-based suggestion is right, it cuts incident resolution time from hours to seconds! How did they do it? 🔄 Two-step approach: ‣ Heuristics (code ownership, directory structure, runtime graphs) reduce thousands of potential changes to a manageable set ‣ Fine-tuned Llama 2 7B ranks the most likely culprits 🎓 Training pipeline: ‣ Continued pre-training on Meta's internal docs and wikis ‣ Supervised fine-tuning on past incident investigations ‣ Training data mimicked real-world constraints (2-20 potential changes per incident) 🔮 Now future developments await: ‣ Language models could handle more of the incident response workflow (runbooks, mitigation, post-mortems) ‣ Improvements in model reasoning should boost accuracy further Read it in full 👉 https://www.tryparity.com/blog/how-meta-uses-llms-to-improve-incident-response

updated a dataset 11 days ago

rbgo/llm-inference-benchmark

liked a Space 25 days ago

lj1995/GPT-SoVITS-v2

View all activity

Organizations

rbgo's activity

upvoted 2 papers about 2 months ago

MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training

Paper • 2403.09611 • Published Mar 14 • 124

MIO: A Foundation Model on Multimodal Tokens

Paper • 2409.17692 • Published Sep 26 • 50

upvoted 10 papers 2 months ago

Scalable AI Safety via Doubly-Efficient Debate

Paper • 2311.14125 • Published Nov 23, 2023 • 2

Towards Bidirectional Human-AI Alignment: A Systematic Review for Clarifications, Framework, and Future Directions

Paper • 2406.09264 • Published Jun 13 • 1

Discovering Language Model Behaviors with Model-Written Evaluations

Paper • 2212.09251 • Published Dec 19, 2022 • 1

Constitutional AI: Harmlessness from AI Feedback

Paper • 2212.08073 • Published Dec 15, 2022 • 2

Training language models to follow instructions with human feedback

Paper • 2203.02155 • Published Mar 4, 2022 • 16

Scaling Laws for Neural Language Models

Paper • 2001.08361 • Published Jan 23, 2020 • 6

Truthful AI: Developing and governing AI that does not lie

Paper • 2110.06674 • Published Oct 13, 2021 • 1

Learning to summarize from human feedback

Paper • 2009.01325 • Published Sep 2, 2020 • 4

The Off-Switch Game

Paper • 1611.08219 • Published Nov 24, 2016 • 1

Concrete Problems in AI Safety

Paper • 1606.06565 • Published Jun 21, 2016 • 1

upvoted an article 3 months ago

Article

Optimizing your LLM in production

Sep 15, 2023

• 15

upvoted an article 5 months ago

Article

Putting RL back in RLHF

Jun 12

• 62

upvoted a paper 6 months ago

Would I Lie To You? Inference Time Alignment of Language Models using Direct Preference Heads

Paper • 2405.20053 • Published May 30 • 2

upvoted 3 collections 7 months ago

Granite Code Models

A series of code models trained by IBM licensed under Apache 2.0 license. We release both the base pretrained and instruct models. • 23 items • Updated 19 days ago • 178

Meta Llama 3

This collection hosts the transformers and original repos of the Meta Llama 3 and Llama Guard 2 releases • 5 items • Updated Sep 25 • 683

WizardLM

0 items • Updated Jul 11 • 103

upvoted a paper 9 months ago

Orca-Math: Unlocking the potential of SLMs in Grade School Math

Paper • 2402.14830 • Published Feb 16 • 24

upvoted a collection 9 months ago

OpenMath

A collection of models and datasets introduced in "OpenMathInstruct-1: A 1.8 Million Math Instruction Tuning Dataset" • 15 items • Updated Oct 1 • 37