NLP PhD student @convai_uiuc | Agents, Reasoning, evaluation etc. https://sagnikmukherjee.github.io https://scholar.google.com/citations?user=v4lvWXoAAAAJ&hl=enhttps://bsky.app/profile/sagnikmukherjee.bsky.socialhttps://bsky.app/profile/sagnikmukherjee.bsky.social/post/3lpnp2u3pgc2y🚨 Paper Alert: “RL Finetunes Small Subnetworks in Large Language Models” From DeepSeek V3 Base to DeepSeek R1 Zero, a whopping 86% of parameters were NOT updated during RL training 😮😮 And this isn’t a one-off. The pattern holds across RL algorithms and models. 🧵A Deep Dive21 May 2025 03:49 +0000at://did:plc:outsnerznqrybvzuoapywek7/app.bsky.feed.post/3lpnp2u3pgc2yhttps://bsky.app/profile/sagnikmukherjee.bsky.social/post/3lom2xer26c2y🚀Our ICML 2025 paper introduces "Premise-Augmented Reasoning Chains" - a structured approach to induce explicit dependencies in reasoning chains. By revealing the dependencies within chains, we significantly improve how LLM reasoning can be verified. 🧵[1/n]07 May 2025 18:52 +0000at://did:plc:outsnerznqrybvzuoapywek7/app.bsky.feed.post/3lom2xer26c2yhttps://bsky.app/profile/sagnikmukherjee.bsky.social/post/3lbihs7gg7s2d📢📢LLMs are biased towards Western Culture. Well, okay, but what do you mean by "Culture"? In our survey of on cultural bias in LLMs, we reviewed ~90 papers. Interestingly, none of these papers define "culture" explicitly. They use “proxies”. [1/7] [Appeared in EMNLP mains]21 Nov 2024 22:03 +0000at://did:plc:outsnerznqrybvzuoapywek7/app.bsky.feed.post/3lbihs7gg7s2d