Leo Boytsov

🧵There is a fundamental issue with reference-based LLM-judges. People implicitly assume a reference-based judge behaves like: score=f(candidate,reference)score=f(candidate,reference) However, the actual behavior is closer to: score=f(candidate,reference,parametric knowledge,pr…

Judging Against the Reference: Uncovering Knowledge-Driven Failures in LLM-Judges on QA Evaluation arxiv.org

This meticulous study delves into the intricate tapestry of lexical biases in LLM-assisted academic writing. It underscores a nuanced interplay of preferred words, unveiling how they have dramatically enhanced and reshaped the realm of science. www.linkedin.com/feed/update/...

LLMs Favorite Words in Academic Literature Dramatically Increase Since GPT Era | Alex Glynn posted on the topic | LinkedIn linkedin.com

🔹Now long-running asynchronous agents expose a weakness in GRPO, so this paper brings the critic back, but with several engineering fixes. 🟦 www.linkedin.com/posts/ravid-...

Just read the new paper from Tsinghua/Z.AI on async RL for agents (arXiv:2607.07508). It comes several weeks after the release of GLM-5.2, in which they mentioned that they use a critic instead of… | Ravid Shwartz Ziv linkedin.com

Interestingly we went full-circle in RL for LLMs and LLM agents: 🔹Initially, OpenAI (and some others) used RLHF with PPO, which requires training a critic (reward) model. 🔹Then, researchers moved from PPO with a critic to critic-free GRPO because critics were expensive and unstable. ↩️

Fun fact: Two some of the most influential statistical NLP papers were authored by Brown et al in 1990 & 2020 1990 Statistical Approach to Machine Translation 2020 Language Models are Few-Shot Learners *) It is not the same Brown **) I believe author names in IBM papers were ordered alphabetically