AI Paper Atlas

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

ACON: Optimizing Context Compression for Long-horizon LLM Agents

Minki Kang, Wei-Ning Chen, Dongge Han, Huseyin A Inan, Lukas Wutschitz, Yanzhi Chen, Robert Sim, Saravan Rajmohan

ICML 2026 · Large language models › Agents, code and math · approach: Long-horizon LLM agents · 1 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 optimized compression guideline (UTCO) for history compressiongpt-4.1 agent with gpt-4.1 compressor · AppWorld test-normalaccuracy: 43.5 -> 56.5Improved
add compression maximization step (CO) after utility maximization (UT) (ablation)ACON history compression · AppWorld test-normalaccuracy: 51.2 -> 56.5Improved
add compression maximization step (CO) after UT (ablation)ACON history compression · OfficeBenchaccuracy: 74.74 -> 72.63Hurt
add compression maximization step (CO) after UT (ablation)ACON history compression · 8-objective QAEM: 0.373 -> 0.335Hurt
use ACON history compression instead of no compressiongpt-4.1 agent · 8-objective QApeak tokens: 10.35 -> 4.71 (10^3)Improved
remove task contrastive feedback from guideline optimization (ablation)ACON UT, o3 optimizer · AppWorldaverage accuracy: 51.2 -> 50.6Hurt
use gpt-4.1 as optimizer instead of o3 (ablation)ACON UT · AppWorldaverage accuracy: 51.2 -> 47.6Hurt
use gpt-5 as optimizer instead of o3 (ablation)ACON UT · AppWorldaverage accuracy: 51.2 -> 50.6Hurt
lower the compression length threshold (ablation)ACON history and observation compression · AppWorld, gpt-4.1accuracy and peak tokensMixed
distill compressor into Qwen3-14B, Qwen3-8B, Phi-4 studentsgpt-4.1 teacher compressor · AppWorld, OfficeBench, 8-objective QAaccuracy: retains over 95% of teacher performanceNo clear effect
use gpt-4.1-mini as compressor without distillation (ablation)ACON history compression · OfficeBench and 8-objective QAaccuracyNo clear effect
add ACON compression for small distilled agentQwen3-14B agent · AppWorldaccuracy (%): 25.6 -> 33.9Improved

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