AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Behavioral Probes for Information Flow in LLM Swarms

Junhui Chen, Yu Zhang, Lantian Li, Dong Wang

NeurIPS 2026 · Large language models › Retrieval and knowledge · approach: Long-context and personalized retrieval · 0 citations (OpenAlex)

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.

Studies LLM swarm interactions through an information-flow lens by introducing behavioral probes that quantify how information is processed, integrated, and relied upon.

Problem

Existing LLM swarm optimization is utility-driven, focusing on final task utility and leaving how information is processed, integrated, and relied upon within individual interactions largely implicit.

Approach

Introduces behavioral probes that quantify information flow patterns in LLM swarms, repurposing External Context (abstract truncated, so the specific probe design is not stated).

Measured effects reported in the paper

What was changedWhereMeasuredEffect
combine ECS and PKS behavioral profiles (Probe-Joint) with utility for node pruningGPT Swarm-style DAG pruning · Random-DAG pruning, Qwen-2.5-32Bredundant LA/BA weight: 2.943 -> 0.997Improved
combine ECS and PKS behavioral profiles (Probe-Joint) with utility for node pruningGPT Swarm-style DAG pruning · Random-DAG pruning, LLaMA-3-70Bredundant LA/BA weight: 2.228 -> 0.498Improved
Probe-Joint instead of GPT Swarm: token consumptionNode pruning · Random-DAG pruning, Qwen-2.5-32Btokens per question: 2294 -> 1589Improved
Probe-Joint instead of utility-only baselines: accuracyNode pruning · Random-DAG pruning, Qwen-2.5-32Baccuracy (%): 80.8 -> 80.0No clear effect
use ECS-only probe instead of joint ECS-PKS (ablation)Probe-guided node pruning · Random-DAG pruning, Qwen-2.5-32Bredundant LA/BA weight: 0.997 -> 2.201Hurt
use PKS-only probe instead of joint ECS-PKS (ablation)Probe-guided node pruning · Random-DAG pruning, LLaMA-3-70Bredundant LA/BA weight: 0.498 -> 2.174Hurt
add PKS-based convergence reward to utility after 50 warm-up iterations (ablation)Reasoning-path search · Taxonomic path search, three warm-startscritical-edge weight growthImproved
remove WebSearch outputs at CombineAnswer stage (ablation)GAIA ToT pipeline · GAIA Level-1PKS: paired mean difference 0.214Improved
remove an entire evidence branch (WebSearch or FileAnalyse) (ablation)GAIA ToT pipeline · GAIA Level-1accuracy (%): 21.2 -> 15.4Hurt
prune only low-ECS evidence (ablation)GAIA ToT pipeline · GAIA Level-1accuracy (%): 21.2 -> 21.2No clear effect
prune high-ECS evidence (ablation)GAIA ToT pipeline · GAIA Level-1accuracy (%): 21.2 -> 15.4Hurt
increase context relevance (bubble to answer)ECS probe, node-level · Kinship questions, Qwen2.5-32BECS (mean adjacent Cliff's delta): 0.624Improved

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