🔥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

Which output is better? [A] or [B]? LLM🤖: B❌ [B] or [A]? LLM🤖: A✅ Thrilled to share our preprint in addressing preference biases in LLM judgments!🧑‍⚖️We introduce ZEPO, a 0-shot prompt optimizer that enhances your LLM evaluators via fairness⚖️ 📰Paper: arxiv.org/abs/2406.11370

Thrilled to share our new paper: "Can LLM be a Personalized Judge?" We investigate the reliability of LLMs in judging user preferences based on personas and propose improvements using verbal uncertainty estimation to enhance accuracy. 🎭👨‍⚖️ 📄 Paper: arxiv.org/abs/2406.11657

Excited to share our latest work: "Jailbreaking as a Reward Misspecification Problem." We explore why safety-aligned LLMs remain vulnerable to adversarial attacks, identifying reward misspecification during the alignment process as a key factor. Find more details in our paper.

Introducing 🪓Segment any Text! 🪓 A new state-of-the-art sentence segmentation tool! Compared to existing tools (and strong LLMs!), our models are far more: 1. efficient ⚡ 2. performant 🔝 3. robust 🚀 4. adaptable 🎯 5. multilingual 🗺

Check out our new reasoning benchmark !!🚀 Is LLM really reasoning or just parroting🦜 ? Why not test if LLM can judge the correctness of different reasoning paths🧠? We cover diverse subjects and reasoning paradigms from logic, coding, maths and more 🔥 📄arxiv.org/abs/2406.13975

First, thanks to the organizers of the workshop. However, we are disappointed about the single brief review that dismissed the attention probing of LLM from the scope, while it is explicitly mentioned in the call for paper. Any response about scope if possible? Ece Takmaz

TRACE the Evidence: Constructing Knowledge-Grounded Reasoning Chains for Retrieval-Augmented Generation (w/ Zaiqiao Meng, Craig Macdonald) Takeaway: using reasoning chains (purely KG triples) built from docs beats using full docs for RAGs. #rag Paper: arxiv.org/pdf/2406.11460

📢New LLM Agents Benchmark! Introducing 🌟MIRAI🌟: A groundbreaking benchmark crafted for evaluating LLM agents in temporal forecasting of international events with tool use and complex reasoning! 📜 Arxiv: arxiv.org/abs/2407.01231 🔗 Project page: mirai-llm.github.io 🧵1/N

Now, you can process 1M context 10x faster, with even better accuracy, try MInference 1.0 right now!

斯坦福爆火的 Prompt 编程框架 DSPy 的 TypeScript 实现来了！ax 实现了 DSPy 支持构建复杂 Agentic Workflow，目前已开源，697 Star 🌟 - 模块化编程：提供标准模块帮助你写 Prompt - 自动编译器：自动为特定 LLM 微调 Prompt 与参数 - 类似 HippoRAG 支持解决复杂多跳检索 github.com/ax-llm/ax

Glad to share two papers accepted to EMNLP 2025 #EMNLP2024 ! One work on improving RAG using reasoning KG chains. w. Jinyuan Fang Craig Macdonald Another is on reducing position bias of LLMs via instruction. w. Meiru Zhang Nigel Collier

🎉Glad to share that our paper "TRACE the Evidence: Constructing Knowledge-Grounded Reasoning Chains for Retrieval-Augmented Generation"(w/ Zaiqiao Meng and Craig Macdonald ) has been accepted at #EMNLP2024 EMNLP 2025 as a findings paper!

💥 Introducing "AutoPSV: Automated Process Supervised Verifier" - accepted at #NeurIPS2024! AutoPSV automatically annotates reasoning steps via confidence tracking, making it efficient and effective even without ground-truth answers. 🔗 arxiv.org/abs/2405.16802 🧵1/5

📢 We are excited to announce the Call for Papers for the 2nd KEIR at ECIR2025 #ECIR2025! 📅 Submission Deadline: January 12, 2025 🔗 Details: keir-ecir2025.github.io Submit your work and join use to explore knowledge-enhanced IR systems! 🚀

If you missed our tutorial, please you can watch recordings via this link: youtube.com/watch?v=Af-_EL…

🚨 New Paper Alert! 🚨 When using LLMs for judgements, ever wondered about the consistency of those judgments? 🤔 Check out our latest work, where we quantify, evaluate, and enhance the logical/preference consistency of LLMs. 📚 🔗 Read more: arxiv.org/abs/2410.02205

📢 Deadline Extension Alert! The deadline for the KEIR workshop has been extended to January 31, 2025 (AOE). Don't miss out this chance to finalise and submit your innovative work . 📜 For detailed submission guidelines, please visit: keir-ecir2025.github.io

Forget just thinking in words. 🚀 New Era of Multimodal Reasoning🚨 🔍 Imagine While Reasoning in Space with MVoT Multimodal Visualization-of-Thought (MVoT) revolutionizes reasoning by generating visual "thoughts" that transform how AI thinks, reasons, and explains itself.

1) Delighted to introduce our latest work🥳 (under review) 🙃 🔗arxiv.org/pdf/2502.15572 🔗 We propose **DReSD: Dense Retrieval for Speculative Decoding** with Huiyin Xue and Gerasimos Lampouras