EngramBench is a new benchmark designed to evaluate skill evolution in autonomous agents by focusing on genuine capability abstraction rather than solution copying. It includes 30 learning tasks and 13 unseen transfer tasks that require agents to manage complex, multi-hour development cycles with LLM‑simulated users. The study shows that while static skill banks cannot eliminate the need for precise code implementation, they effectively reduce redundant context and cut overall coding time by more than 55%.
By Zhixuan Tan, Pengjie Gu, Zhao Li, Yihan Hu, Xu He, Dong Li, Jianye Hao
arXiv:2606. 10087v1 Announce Type: cross Abstract: Pre-training on raw code teaches syntax but provides sparse signal for diverse real-world task formats.
By Ankit Gupta, Aditya Prasad, Rameswar Panda
The paper introduces a self‑evolving harness framework where a frozen language‑model agent first solves tasks and then edits its own harness based on run records. Using a 49‑line seed harness, the evolved harness improves average scores on in‑distribution benchmarks by 4.48 points and on out‑of‑distribution benchmarks by 12.64 points, surpassing Codex on the former and matching it on the latter. Continued evolution on a specific out‑of‑distribution benchmark further raises performance, and the study analyzes emergent mechanisms such as output truncation and history compaction.
By Qiankai Xu
arXiv:2609.08435v2 Announce Type: new
Abstract: In persistent interactions, long contexts may encode an evolving process rather than a fixed record: later events can revise or revoke earlier informat...
By Ziliang Zhao, Zenan Xu, Shuting Wang, Zhao Wang, Bowen Cao, Minda Hu, Lincheng Li, Pluto Zhou, Zhicheng Dou
arXiv:2605. 28556v2 Announce Type: replace Abstract: As agent capabilities advance, existing benchmarks, such as $\tau^2$-Bench, are becoming increasingly saturated.
By Tomer Keren, Nitay Calderon, Asaf Yehudai, Yotam Perlitz, Michal Shmueli-Scheuer, Roi Reichart
LongWoF-Bench is a new benchmark of 778 machine‑verifiable long‑workflow tasks spanning code generation, agent‑environment synthesis, mathematical reasoning, and rule following. The study shows that EvoMap Genes—structured representations of verifier‑confirmed execution trajectories—outperform the Skill baseline by 8.7–15.5 percentage points across seven models, and for Claude Opus they enable 39 additional task completions while cutting token consumption by 9.9%. The results demonstrate that verified execution experience can be externalized and reused, improving long‑workflow completion without repeatedly discovering new strategies.
By Xiao Zhang, Qumeng Sun, Jihao Li, Yiming Ren, Xiang Liu, Haoyang Zhang, Junjie Wang