🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Given its promising results in reasoning-oriented tasks, how can we leverage RL with tool-specific reward functions for improved tool learning? How is the performance of smart rewards compared to SFT? Does it also lead to better thinking skills in tool-calling tasks? Introducing

This is an important milestone for enabling LLM-based agents. Reward is all you need for Tool Learning! GRPO achieves significant improvements over base and SFT models for BFCL v3, API-Bank and Bamboogle agentic benchmark tasks. Congratulations Emre Can Acikgoz and Cheng Qian

What do we want from "Conversational Agents" on top of language agents? What is missing in the current Conversational Agent systems? Here is our desideratum with a comprehensive survey of the recent advances in the field. Bonus is a live github collection of new papers organized

🚀🤖 Check out Emre's cool new survey on conversational agents with live github collection of new papers organized in an easy to follow taxonomy!

🚀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]

Checkout this new work by Sagnik Mukherjee on premise augmented reasoning chains!

super excited to see yourbench converge to be the default generative benchmarking / synthetic data creation solution for llms 💛

Check out Beyza's awesome survey on LLM persuasion 🤖 with clear definitions of each persuasive aspect and key challenges and future directions 🎉

Excited to share our survey on computational persuasion - check it out to learn more about 🤖 AI as Persuader, 🎯AI as Persuadee, and ⚖️AI as Persuasion Judge!

A must read paper by Beyza Bozdag and others! 😀 ConvAI@UIUC

Congratulations Emre 🎉🎉 See you in Vienna 🇦🇹

🚨 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 Dive

Paper - arxiv.org/abs/2505.11711 Work done with amazing collaborator Lifan Yuan and advised by our amazing advisors dilek hakkani-tur and Hao Peng

Would models know more about Indian food in Hindi and Turkey’s history in Turkish? Does the language of a question affect an LLM’s answer? ✨Yes!✨ Ishika Agarwal and I are excited to announce our newest preprint in which we explore “Language Specific Knowledge (LSK)”.

Would models know more about Indian food in Hindi and Turkey’s history in Turkish? Does the language of a question affect an LLM’s answer? ✨Yes!✨ Beyza Bozdag and I are excited to announce our newest preprint in which we explore “Language Specific Knowledge (LSK)”.

[7/7] We are grateful to our advisor dilek hakkani-tur, special shoutout to Sagnik Mukherjee and Priyanka Kargupta, and other ConvAI@UIUC lab members 🎉 📄Paper: arxiv.org/pdf/2505.14990 💻Code: github.com/agarwalishika/…

[7/7] We are grateful to our advisor dilek hakkani-tur, special shoutout to Sagnik Mukherjee and Priyanka Kargupta, and other ConvAI@UIUC lab members 🎉 📄Paper: arxiv.org/pdf/2505.14990 💻Code: github.com/agarwalishika/…

Reinforcement Learning Finetunes Small Subnetworks in Large Language Models Author's Explanation: x.com/saagnikkk/stat… Overview: Reinforcement learning substantially improves LLMs through an intrinsic phenomenon of parameter update sparsity, where only 5-30% of parameters

Thanks for covering our work !

Did you know that papers from the journal Computational Linguistics will also be featured at ACL 2025? 🤩 This means even more cutting-edge NLP research to explore in Vienna! Don't miss out! #ACL2025NLP #ComputationalLinguistics #NLProc 2025.aclweb.org/program/cl_pap…