pith. sign in
Pith Number

pith:3AKEZE3T

pith:2025:3AKEZE3TAPIIHZYC5RBY7ZDHZN
not attested not anchored not stored refs resolved

MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent

Hao Zhou, Hongli Yu, Jiangjie Chen, Jiangtao Feng, Jingjing Liu, Mingxuan Wang, Qiying Yu, Tinghong Chen, Weinan Dai, Wei-Ying Ma, Ya-Qin Zhang

MemAgent lets LLMs handle millions of tokens by segmenting input and overwriting memory after RL training on 32K texts.

arxiv:2507.02259 v1 · 2025-07-03 · cs.CL · cs.AI · cs.LG

Add to your LaTeX paper
\usepackage{pith}
\pithnumber{3AKEZE3TAPIIHZYC5RBY7ZDHZN}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp
2 Internet Archive
3 Author claim open · sign in to claim
4 Citations open
5 Replications open
Portable graph bundle live · download bundle · merged state
The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

MemAgent has demonstrated superb long-context capabilities, being able to extrapolate from an 8K context trained on 32K text to a 3.5M QA task with performance loss < 5% and achieves 95%+ in 512K RULER test.

C2weakest assumption

That the overwrite memory strategy combined with multi-conversation RL training will continue to prevent performance degradation when scaling far beyond the 32K training length without additional mechanisms or data.

C3one line summary

MemAgent uses multi-conversation RL to train a memory agent that reads text in segments and overwrites memory, extrapolating from 8K training to 3.5M token QA with under 5% loss and 95%+ on 512K RULER.

References

65 extracted · 65 resolved · 33 Pith anchors

[1] RULER: What's the Real Context Size of Your Long-Context Language Models? 2024 · arXiv:2404.06654
[2] HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering 2018 · arXiv:1809.09600
[3] Learning to reason with llms 2024
[4] Gemini 2.0 flash thinking, 2024 2024
[5] Grok 3 beta — the age of reasoning agents, 2024 2024

Formal links

2 machine-checked theorem links

Cited by

37 papers in Pith

Receipt and verification
First computed 2026-05-17T23:38:52.698861Z
Builder pith-number-builder-2026-05-17-v1
Signature Pith Ed25519 (pith-v1-2026-05) · public key
Schema pith-number/v1.0

Canonical hash

d8144c937303d083e702ec438fe467cb66da96ccf76f3bf79fd0a307195bf4b8

Aliases

arxiv: 2507.02259 · arxiv_version: 2507.02259v1 · doi: 10.48550/arxiv.2507.02259 · pith_short_12: 3AKEZE3TAPII · pith_short_16: 3AKEZE3TAPIIHZYC · pith_short_8: 3AKEZE3T
Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/3AKEZE3TAPIIHZYC5RBY7ZDHZN \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: d8144c937303d083e702ec438fe467cb66da96ccf76f3bf79fd0a307195bf4b8
Canonical record JSON
{
  "metadata": {
    "abstract_canon_sha256": "909f67bdf57361e1e01dbde35b145d65f5a3e5e6d9083acc5b7b887a65a2b28d",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.LG"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2025-07-03T03:11:50Z",
    "title_canon_sha256": "1c5f2eb4f7d03e2225f06250d7bc510eab14efdecea0d98d161a35291af95019"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2507.02259",
    "kind": "arxiv",
    "version": 1
  }
}