AI Paper Atlas

Research map / Multi-agent systems and games / LLM agents

IterResearch: Rethinking Long-Horizon Agents via Markovian State Reconstruction

Guoxin Chen, Zile Qiao, Xuanzhong Chen, Donglei Yu, Haotian Xu, Xin Zhao, Ruihua Song, Wenbiao Yin, Huifeng Yin, Liwen Zhang, Kuan Li, Minpeng Liao and 4 more

ICLR 2026 · Multi-agent systems and games › LLM agents · approach: Robust and efficient agent execution

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes IterResearch, a deep-research agent that replaces a single ever-expanding context with iterative Markovian state reconstruction to handle long-horizon tasks.

Problem

Deep-research agents that accumulate all information in one expanding context window suffer from context suffocation and noise contamination on long-horizon tasks.

Approach

Replaces the mono-contextual paradigm with iterative rounds in which the agent reconstructs a compact Markovian state from its previous state rather than appending everything to one growing context.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use iterative workspace reconstruction instead of mono-contextual accumulation (same data, mono gets 64K vs 40K context) (ablation)IterResearch agent · six deep-research benchmarks (avg)accuracy: 36.5 -> 49.1 average (BC 25.4 -> 37.3)Improved
add efficiency-aware geometric discounted reward (EAPO) over GSPO (ablation)IterResearch · six benchmarks avgaccuracy and avg turns: 48.3 -> 49.1 accuracy; 19.13 -> 18.04 turnsImproved
add RL (GSPO/EAPO) after SFT (ablation)IterResearch · six benchmarks avgaccuracy: 45.5 -> 49.1Improved
mix iterative-paradigm trajectories into mono-agent training data (ablation)Mono-Agent · six benchmarks avgaccuracy: 36.5 -> 41.9Improved
raise maximum interaction turns from 2 to 2048 (ablation)IterResearch · BrowseComp (200 subset)accuracy: 3.5 -> 42.5 (text says 5.5% -> 50.1%, disagreeing with figure)Improved
apply IterResearch paradigm as prompt instead of ReActo3 · BrowseCompaccuracy: 49.7 -> 62.4Improved
apply IterResearch paradigm as prompt instead of ReActDeepSeek-V3.1 · BrowseCompaccuracy: 30.0 -> 49.2Improved
EAPO vs GSPO on BrowseComp (ablation)IterResearch · BrowseCompaccuracy: 38.3 -> 37.3No clear effect

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