Sentry is a failure‑management layer for large language model agents that learns from failures at test time. It retrieves relevant lessons from an external playbook when a failure occurs, verifies recovery without task rewards, and stores new lessons only if recovery succeeds, keeping the playbook out of the agent’s context. Across multiple benchmarks, Sentry outperforms both runtime‑intervention and context‑evolution baselines, and its lessons transfer to unseen tasks.
By Changxiu Ji, Amy Lu, Qizheng Zhang, Kunle Olukotun
The paper re‑evaluates memory‑based self‑improving agents by adding multiple runs to measure variance and by randomizing task order. It finds that agent performance is noisy in complex, multi‑step environments and that improvement depends heavily on the sequence of tasks, revealing a hidden curriculum effect. The authors suggest that underspecification of tasks and environments contributes to this fragility and demonstrate that adding detailed rubrics and feedback can partially mitigate performance drops, though gaps remain.
By Qinyuan Ye, Yu Li, Yada Pruksachatkun, Jiaxin Zhang, Chien-Sheng Wu
The paper re‑evaluates memory‑based self‑improving agents by running multiple trials and shuffling task orders, revealing that agent performance is noisy and highly sensitive to task sequencing. It shows that implicit curricula in default task orders act as hidden prerequisites for success, and that underspecification of tasks and environments contributes to fragility. Adding detailed rubrics and environment feedback partially mitigates performance drops but significant gaps remain, underscoring the need for stricter evaluation protocols and better human oversight.
arXiv:2608.22533v1 Announce Type: new
Abstract: Autonomous computer-use agents are increasingly applied to long-horizon tasks requiring coordinated application calls, persistent state tracking, and v...
By Zheyuan Deng, Binghang Lu, Hanqi Feng, Shirley Huang, Dianzhuo Wang, Yuanda Xu, Zhiwei Zhang, Yige Sun, Changhong Mou, Runyu Zhang, Yuexing Hao, Barnabas Poczos, Xiaomin Li
AdaMem introduces adaptive memory policies that allow personalized agents to decide what information to write into long‑term memory based on user preferences for each interaction context. Each policy is updated from periodic feedback and controls subsequent memory writing, aiming to improve relevance and reduce unnecessary memory persistence. In experiments on AdaMem‑Bench, AdaMem raises QA accuracy from 80.0% to 84.35% while cutting persistent memory by 9.27%, though models still struggle to execute policies reliably.
By Xingyu Chen, Rui Wang, Zhaopeng Tu, Liefeng Bo
Autonomous computer-use agents are increasingly applied to long-horizon tasks requiring coordinated application calls, persistent state tracking, and verifier-sensitive writes, yet they remain prone t...
Learn2Play Bench is a new benchmark that tests how well large language model agents learn from experience in unfamiliar, text‑based games with novel or counterintuitive rules. The benchmark provides reproducible feedback, automatic scoring, and varied game instances to evaluate learning across repeated attempts and transfer to new situations. Findings show that retaining full action records aids learning, human players outperform agents, and the choice of harness significantly impacts performance and inference cost.
By Yibo Li, Jinhang Qiu, Zhi Zheng, Qianyun Guo, Jiaying Wu, Shuo Ji, Bryan Hooi
arXiv:2607.10526v5 Announce Type: replace
Abstract: Self-improving personal agents write profiles, memories, and reusable skills that carry over from one chat to the next. Prior works ask whether a m...
By Xutao Mao, Liangjie Zhao, Leyao Wang, Rui Qian, Qiang Huang, Wentao Wang, Bo Han, Xiang Zheng, Cong Wang
The paper investigates how language models can aid their own continual learning by generating replay data and selecting preservation targets. By organizing task descriptions into Task Banks, the authors show that replay generated by frozen models reduces forgetting on tasks outside the bank and that informed selection of replay targets can improve retention. The study demonstrates that models can support broader self‑preservation, especially when larger models are used, and that replacing new‑task examples with replay can enhance preservation under a fixed example budget.
By Dongkyu Cho, Rumi Chunara, Ali Behrouz
Memento 3 is a frozen large‑language‑model agent that learns explicit world models by maintaining a natural‑language rulebook in external memory. The agent iteratively observes, reflects, revises the rulebook, compiles it into executable code, and verifies updates through prediction error and replay, accepting only code that the LLM deems faithful to the rulebook. In experiments, a single‑model version solved all 25 public ARC‑AGI‑3 games with perfect efficiency, and a learned feedback controller achieved perfect scores in Atari Pong without further LLM calls.
By Haoyu Zhao, Zhengxu Yu, Zhiyuan He, Meng Fang, Rasul Tutunov, Haitham Bou-Ammar, Weilin Luo, Jun Wang
arXiv:2610.14495v1 Announce Type: new
Abstract: Explainer videos -- a worked geometry problem, why the sky is blue, how a dynasty fell -- are usually one more target for pixel video generation. We ar...
By Wei Lin
The paper introduces Just-in-Time Memory (JitMem), a system that defers memory curation until a task is read, allowing a curator to synthesize task‑specific memory payloads based on the current query. Unlike traditional write‑time curation, JitMem retains raw trajectories and trains the curator using immediate task success, avoiding long‑horizon credit‑assignment issues. Experiments on ALFWorld, WebShop, and τ²‑bench show JitMem consistently outperforms both no‑memory agents and existing write‑time memory methods, with improvements of up to 16.3 absolute success‑rate points.
whyItMatters":"By curating memory at read time, JitMem enables more effective, task‑adaptive recall that directly improves agent performance across diverse benchmarks."
By Yefan Zhou, Yang Li, Zeyu Leo Liu, Semih Yavuz, Shafiq Joty