Baishakhi Ray (@baishakhir) 's Twitter Profile
Baishakhi Ray

@baishakhir

Associate Professor, Columbia

ID: 1138838376

linkhttp://rayb.info calendar_today01-02-2013 06:05:29

145 Tweet

817 Followers

379 Following

Percy Liang (@percyliang) 's Twitter Profile Photo

We should call models like Llama 3, Mixtral, etc. “open-weight models”, not “open-source models”. For a model to be open-source, the code and training data need to be public (good examples: GPT-J, OLMo, RedPajama, StarCoder, K2, etc.). Weights are like an exe file, which would be

Baishakhi Ray (@baishakhir) 's Twitter Profile Photo

We need to promote open source culture for both models and data. This will also require a repository management for data to track its evolution.

Abhik Roychoudhury (@abhikroychoudh1) 's Twitter Profile Photo

CodeRover++, new version of AutoCodeRover, is here! A pragmatic outlook to autonomous software engineering of the future ! Optimising for multiple objectives (efficacy, cost and time), while automatically solving software engineering tasks. Future Large Language Model (LLM)

Baishakhi Ray (@baishakhir) 's Twitter Profile Photo

A new dataset and some potential remedies to reduce API hallucinations of LLMs. Our initial studies found that even SOTA LLMs like GPT4o hallucinates significantly for evolving APIs.

ASE 2024 (@ase_conf) 's Twitter Profile Photo

🏆The Distinguished Papers for ASE 2024 have been announced! Discover groundbreaking research and exceptional contributions that are shaping the future of software engineering. Congrats to all the authors! Explore the full list at the link below: conf.researchr.org/info/ase-2024/…

Alex Reibman 🖇️ (@alexreibman) 's Twitter Profile Photo

OpenAI’s biggest rival is shaking things up. Anthropic invited 200+ elite hackers to their SF headquarters to see what’s possible with Claude Here’s what we saw at the Anthropic x Menlo Ventures Builder Day Hackathon (🧵):

OpenAI’s biggest rival is shaking things up.

Anthropic invited 200+ elite hackers to their SF headquarters to see what’s possible with Claude

Here’s what we saw at the <a href="/AnthropicAI/">Anthropic</a> x <a href="/MenloVentures/">Menlo Ventures</a> Builder Day Hackathon (🧵):
Baishakhi Ray (@baishakhir) 's Twitter Profile Photo

Updating real-world large legacy projects like binutils? Meet (arxiv.org/abs/2501.14257) C2SaferRust: leveraging program analysis & LLMs to create idiomatic, safer Rust with (↓38%) raw pointers & (↓28%) unsafe code while preserving functionality 🚀 #rustlang #AI4code #AIAgent

Baishakhi Ray (@baishakhir) 's Twitter Profile Photo

I believe that detecting misbehavior in CoT and fixing it—at least partially through automation—would be a highly effective way to enhance the model's performance. This aligns with the idea that clean data significantly improves model quality.

Pieter Abbeel (@pabbeel) 's Twitter Profile Photo

Founders who were PhD or post-doc in my lab at Berkeley, **largely funded by NSF / DoD grants**, start-up, market cap (collected by OpenAI Deep Research)

Founders who were PhD or post-doc in my lab at Berkeley, **largely funded by NSF / DoD grants**, start-up, market cap (collected by OpenAI Deep Research)
Baishakhi Ray (@baishakhir) 's Twitter Profile Photo

Our empirical study on the impact of library evolution on the LLM’s code generation capability. Sachit Kuhar will present the paper tomorrow at NAACL. If you are around please attend the talk.