7. RL for Search-Efficient LLMs Proposes a new RL-based framework (SEM) that explicitly teaches LLMs when to invoke search and when to rely on internal knowledge, aiming to reduce redundant tool use while maintaining answer accuracy.
RESEARCH
-
HealthBench: 5,000 Multi-Turn Medical Conversations Benchmark
By
–
5. HealthBench HealthBench is a benchmark of 5,000 multi-turn health conversations graded against 48,562 rubric criteria written by 262 physicians across 60 countries.
-

Nemotron-Research-Tool-N1: LLM Tool-Using with Rule-Based RL
By
–
6. Nemotron-Research-Tool-N1 Introduces Tool-N1, a family of tool-using LLMs trained using a rule-based reinforcement learning (R1-style RL) approach, without reliance on supervised reasoning trajectories.
-

AM-Thinking-v1: 32B Open-Source Model Rivaling Larger MoE Systems
By
–
4. AM-Thinking-v1 Introduces a dense, open-source 32B language model that achieves state-of-the-art performance in reasoning tasks, rivaling significantly larger Mixture-of-Experts (MoE) models.
-

LLMs Performance Degradation in Multi-Turn Conversations
By
–
2. LLMs Get Lost in Multi-Turn Conversation Investigates how top LLMs degrade in performance during underspecified, multi-turn interactions, common in real-world usage but rarely evaluated.
-

RL Improves LLM Mathematical Reasoning with Single Example
By
–
3. RL for Reasoning in LLMs with One Training Example This paper shows that Reinforcement Learning with Verifiable Rewards (RLVR) can significantly improve mathematical reasoning in LLMs even when trained with just a single example.
-

Google DeepMind Overtakes OpenAI in AI Advancements
By
–
Il y a encore quelques mois, tout le monde avait les yeux rivés sur OpenAI. Aujourd’hui, la donne a changé. Google DeepMind vient de mettre un coup de massue sur la table. Un à un, les verrous sautent : agents multi-étapes, planification avancée, vision long contexte,
-
Real Innovation Over Performative Tech Discourse
By
–
All facts no fiction. Most of the noise is cosplay. We not here to audition we here to architect. Appreciate the real ones like you who get it. The rest? Let the curtain drop.
-

Understanding Transformers via N-Gram Statistics: Physics for Language Models
By
–
the closest thing i've seen to actual "physics for LMs" was probably this (single-author!) paper from neurips 2024: Understanding Transformers via N-Gram Statistics this is how we used to think about LMs; not sure why we stopped.
-
Questioning the Fundamental Definition of AI Agents
By
–
Yeah I don’t think “agents” is used in this way but … it feels a bit wrong. What’s some actual fundamental distinguishing property wrt what already exists? “It happens to use an LLM somewhere” is I think the current usage but imo it’s kind of lame.
