A paper submitted to arXiv on 29 Sep 2026 introduces Context Language Models (CLMs), language models that natively manage their own context by treating the context as a file the model can update witho…
A September 22, 2026 arXiv paper by Laizhen Li introduces Growing Harness, a failure-guided training paradigm that learns an LLM agent's harness as reusable executable code from a strategy-free scaffo…
A first-move retrieval module called Question's Gambit raised downstream agent answer accuracy from 83.1% to 90.5% with gpt-5.5 over the Pi-Serini agentic baseline on the BrowseComp-Plus benchmark, ac…
A context gate that filters retrieved documents before they enter an agent's context cuts input tokens by 1.4× with no accuracy loss on the BrowseComp-Plus benchmark, but doubles per-query latency and…
TRACE, a novel dense credit-assignment framework for reinforcement learning, dramatically improved agent performance on the BrowseComp-Plus benchmark, boosting Qwen3-4B from 7.2 to 35.6 and Qwen3-30B-…
ECHO, a novel framework for traceable context reconstruction in agentic reinforcement learning, achieves 43.4% held-out accuracy on BrowseComp-Plus, outperforming GRPO's 28.9% and SUPO's 36.1% with fe…
Researchers introduced MosaicLeaks, a benchmark revealing that deep research agents leak private information through web queries, and proposed Privacy-Aware Deep Research (PA-DR) training that reduced…
ARES (Adaptive Reasoning Effort Selection) is a framework that reduces the cost of multi-step LLM agent tasks by using a lightweight router to predict the minimum reasoning effort needed for each step…