Backend
Surfacing Disagreement Beats Resolving It: A Gated Ablation Study of Provenance-Weighted Memory for Multi-Agent Debate
arXiv cs.MA0 views
arXiv:2611.22461v1 Announce Type: new Abstract: Shared memory for multi-agent debate is usually built to resolve inconsistency: retrieval returns one winning fact and the disagreement disappears before any agent sees it. We ask whether that is correct. In DebateGraph, four backends share one contract (read/write/snapshot), so five mechanisms from the agent-memory literature can be removed independently under a fixed budget, each gated on a pre-declared threshold. Across 39 runs one survives: preserving contradictions and passing the ranked conflict set to the debate, rather than collapsing it to a winner, improves F1 by +0.340 (95% CI [+0.267, +0.414]) and cuts attack success from 22 to 7 of 122 poisoned items (McNemar exact p=0.0003), at equal cost. Reaching it meant fixing a bottleneck in our harness: conflict sets are built only over retrieved edges, and our pre-declared cap left three quarters of planted contradictions unretrievable, first measuring the mechanism at +0.084. The effect replicates on two retrieval substrates and directionally on HotpotQA documents. No other mechanism survives, and two are worse than inert: once contradictions reach the prompt, a write trust-gate and the router's provenance features each cost accuracy (-0.075 and -0.096 F1). The gate fails structurally: poisoned objects are always entities the graph already contains, against 53% of gold, so groundedness scores the lie above a truthful first mention. We release the harness, attack set, and logs, and argue that silently reconciling memory is the assumption worth abandoning.
Source:https://arxiv.org/abs/2611.22461PreviousEmergence and Collapse of Leader-Follower Flocks through Phantom Targets: Surprise Nullification in Bayesian and Inverse Bayesian InferenceNextFieldGuide: A Corpus Map for Token-Efficient Agentic Document Exploration
Related posts
Clash can't reach your LAN? Three cases, three different fixes
0dev.to Linux
When an Empty String Makes a Python CLI Read from stdin
0dev.to Python

Measuring 'unblur' tools against ground truth: sharper isn't closer
0dev.to Python
MySQL 1226 max_connections_per_hour on Shared Hosting
0dev.to Python
Beyond the Black Box: Building a Scalable, Stem-Aware Audio Engine for the Web
0dev.to Python
🚀 Galactic Express: How Python Loops Saved the Universe!
0dev.to Python