Paper Archive

输入关键词搜索 114 项内容

tracked author

Nusrat Jahan Lia

LRE first author and owner of its public code repository. Her homepage records University of Dhaka software-engineering training, a current Fatima Institute research fellowship, and a broader research line spanning LLM-agent reliability, memory, and multilingual NLP.

1 archived notes X: not-found HomepageGitHub

Representative Papers

来自作者已核验个人主页的重点论文；本站单篇归档见下方 Related Notes。

  1. 01 AgentCheck: A Reproduce–Intervene–Mitigate Workbench for LLM Agents over MCP
  2. 02 Learning What Not to Forget: Long-Horizon Agent Memory from a Few Kilobytes of Learning
  3. 03 AgentCollabBench: Diagnosing When Good Agents Make Bad Collaborators
  4. 04 Cross-Lingual Sentiment Misalignment: Auditing Multilingual Language Models for Inversion Risk, Dialectal Representation, and Affective Stability
  5. 05 Read Between the Lines: A Benchmark for Uncovering Political Bias in Bangla News Articles
  6. 06 Exploring Cross-Lingual Knowledge Transfer via Transliteration-Based MLM Fine-Tuning for Critically Low-resource Chakma Language
  7. 07 Adult Attitudes about School Smartphone Bans: A Global Survey of 35 Countries

Related Notes

按论文归档时间排序，展示该作者在本站已经出现的材料。

归档

Learning What Not to Forget: Long Horizon Agent Memory from a Few Kilobytes of Learning

LRE 用日志中的未来复用信号训练轻量逻辑回归，对历史单元做查询无关打分并按预算逐字保留；单随机种子 AppWorld 中以零压缩器调用得到 41.1% 任务目标完成率、接近全历史的 44.0%，LoCoMo 上也是最佳受预算策略，但 Hard 与 LongMemEval 结果显示它尚未跨域占优，2048 token 只约束较旧历史。

待审阅 2606.20954-lre-learned-relevance-eviction Agent MemoryLong ContextAgent Workflow