AI Paper Atlas

Research map / Large language models / Agents, code and math

ParamMem: Augmenting Language Agents with Parametric Reflective Memory

Tianjun Yao, Yongqiang Chen, Yujia Zheng, Pan Li, Zhiqiang Shen, Kun Zhang

ICML 2026 · Large language models › Agents, code and math · approach: Long-horizon LLM agents · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add ParamMem parametric memory to Reflexion (ParamAgent)Llama-3.1-8B agent · HumanEvalPass@1: 76.22 -> 82.93Improved
add ParamMem to Reflexion (ParamAgent)Llama-3.1-8B agent · 2WikiMultiHopQAaccuracy: 78.67 -> 88.67Improved
add ParamMem to DoT-bank cross-sample memory (ParamAgent-plus)Llama-3.1-8B agent · MATHaccuracy: 73.02 -> 75.45Improved
use diversity-driven ParamMem instead of accuracy-driven Retroformer moduleLlama-3.1-8B agent · MBPPPass@1: 42.82 -> 67.00Improved
use Retroformer rather than ParamMem on mathMistral-7B-v0.2 agent · MATHaccuracy: 43.53 -> 28.06Hurt
train ParamMem on data generated by Llama-3.1-8B itself (self-improvement)ParamAgent · HumanEvalPass@1: 73.17 -> 78.05Improved
train ParamMem on self-generated data, plus cross-sample memoryParamAgent-plus · HotpotQAaccuracy: 72.00 -> 83.33Improved
add Llama-3.1-8B-based ParamMem to a Qwen3-Next-80B agentParamAgent-plus · LiveCodeBenchPass@1: 62.00 -> 63.33Improved
use Qwen3-Next-30B ParamMem instead of Llama-3.1-8B ParamMem for coding (ablation)ParamAgent-plus on Qwen3-Next-80B · LiveCodeBenchPass@1: 63.33 -> 68.00Improved
use Qwen3-Next-30B ParamMem instead of Llama-3.1-8B on QA (ablation)ParamAgent-plus on Qwen3-Next-80B · HotpotQAaccuracy: 85.00 -> 83.67Hurt
train ParamMem on 500 clustered samples instead of 8000 (ablation)ParamAgent · HumanEvalPass@1: 82.93 -> 81.71No clear effect
add cross-sample memory to ParamAgent trained on 500 samples (ablation)ParamAgent-plus · HumanEvalPass@1: 81.71 -> 86.59Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy