How this thread developed

1. June 2026

   Pruning context to recent tool calls improves agent reliability and efficiency

   Pruning context to recent tool calls improves reliability and efficiency, comparing four context engineering strategies.

2. June 2026

   An agentic framework that fills in missing context for real-world image generation

   Agentic framework that fills in missing context for real-world image generation via Context-Aware Planning.

3. 14 further papers

   Sept 2026 · University of California San Diego

   Coding agents retrieve better by targeting missing evidence sets

   Proposes MSS-Complement retrieval that targets missing evidence, directly improving context construction for coding agents.

4. 1 further paper

   Sept 2026 · Southern University of Science and Technology, Peking University

   Selective attention calls, guided by model's own state, speed long-context inference

   ODA uses a trained recall head to selectively invoke global attention during long-context inference, directly optimizing context access efficiency.

5. Sept 2026 · NLPR&MAIS, Institute of Automation, Chinese Academy of Sciences

   Bounded visual workspaces improve multimodal agents’ accuracy and efficiency

   Bounded visual workspace via ledger storage reduces context length and improves accuracy.

6. Sept 2026 · Carnegie Mellon University, Bosch Center for AI

   Dropping Rather Than Rewriting Context Cuts Long-Horizon Agent Costs

   Proposes a context-compaction method that truncates rather than rewrites, reducing errors and costs in long-horizon agent tasks.

   released code

7. Sept 2026 · Shanghai Jiao Tong University, Theseus Labs

   Restructuring agent environments improves performance on noisy, evolving tasks

   Introduces Env-Rethink, which restructures agent environment information to improve performance on noisy tasks.

8. Sept 2026 · TierFlow Team, Renmin University of China

   Agents Can Drop Old Reasoning After Externalizing Task State

   Shows that agents can drop old reasoning after externalizing task state using online compression, reducing token use and improving accuracy on WorkBuddyBench.

8 of 23 papers shown