Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Detecting Perspective Shifts in Multi-Agent Systems
Eric W Bridgeford, Hayden Helm
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use TDKPS embeddings with agent-level permutation test instead of DCorr | agent-level change detection (simulation) · Temporal Gaussian blobs, 5 agents, effect size 0.2 | power: DCorr power <0.15; TDKPS nearly quadruples it | Improved |
| increase number of queries (ablation) | DCorr agent-level test · Temporal Gaussian blobs | Type I error | Hurt |
| increase number of response replicates (ablation) | TDKPS agent-level test · Temporal Gaussian blobs | Type I error rate: 5--10% inflation decreases with more replicates | Improved |
| use sample-splitting variant TDKPS-C (ablation) | TDKPS agent-level test · Temporal Gaussian blobs | Type I error and power | Mixed |
| use TDKPS embeddings with paired energy test instead of DCorr or DCorr o TDKPS | group-level change detection (simulation) · Temporal Gaussian blobs, SNR<<0.02 | power | Improved |
| increase agents per group from 10 to 80 (ablation) | PE o TDKPS · Temporal Gaussian blobs, effect size 0.2 | power: ~50% at 10 agents -> >95% at 80 agents | Improved |
| use public health queries instead of orthogonal queries | TDKPS agent-level test on congressperson agents · COVID-19 natural experiment | Kendall tau with proximity to COVID-19: 0.01 -> 0.51 | Improved |
| use public health queries instead of general political queries | TDKPS agent-level test on congressperson agents · COVID-19 natural experiment | Kendall tau with proximity to COVID-19: 0.34 -> 0.51 | Improved |
| change embedding model (four architectures) (ablation) | TDKPS temporal shift pattern · congressperson agents | Spearman rho of shift pattern: rho > 0.99 | No clear effect |
| change retrieval depth (ablation) | TDKPS temporal shift pattern · congressperson agents | Spearman rho of shift pattern: rho > 0.81 | No clear effect |
Closest papers by meaning
- When Agents “Misremember” Collectively: Exploring the Mandela Effect in LLM-based Multi-Agent Systems (ICLR 2026 · similarity 0.86)
- ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memory (ICLR 2026 · similarity 0.85)
- G-Memory: Tracing Hierarchical Memory for Multi-Agent Systems (NeurIPS 2025 · similarity 0.85)
- Mental-Models for Multi-Agent Systems (NeurIPS 2026 · similarity 0.85)
- Test-Time Adaptation for LLM Agents via Environment Interaction (ICLR 2026 · similarity 0.85)
- Emergent Coordination in Multi-Agent Language Models (ICLR 2026 · similarity 0.85)
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.
