# What Confidence Routing Is Actually Doing: Auditing Routing, Calibration, and Commitment in Multi-Agent Deliberation

arXiv: https://arxiv.org/abs/2609.27822 · HTML: https://mlsanity.com/paper/2609.27822/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 32 references that are in the map.
- First posted on arXiv 18 Aug 2026.

## Builds on (most cited first, 19 of 19)

- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters](https://mlsanity.com/paper/2408.03314/) (2024)
- [Improving Factuality and Reasoning in Language Models through Multiagent Debate](https://mlsanity.com/paper/2305.14325/) (2023)
- [Model Cards for Model Reporting](https://mlsanity.com/paper/1810.03993/) (2018)
- [Large Language Models Cannot Self-Correct Reasoning Yet](https://mlsanity.com/paper/2310.01798/) (2023)
- [Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs](https://mlsanity.com/paper/2306.13063/) (2023)
- [Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback](https://mlsanity.com/paper/2305.14975/) (2023)
- [Teaching Models to Express Their Uncertainty in Words](https://mlsanity.com/paper/2205.14334/) (2022)
- [Mixture-of-Agents Enhances Large Language Model Capabilities](https://mlsanity.com/paper/2406.04692/) (2024)
- [A Dynamic LLM-Powered Agent Network for Task-Oriented Agent Collaboration](https://mlsanity.com/paper/2310.02170/) (2023)
- [Debating with More Persuasive LLMs Leads to More Truthful Answers](https://mlsanity.com/paper/2402.06782/) (2024)
- [Self-Preference Bias in LLM-as-a-Judge](https://mlsanity.com/paper/2410.21819/) (2024)
- [Should we be going MAD? A Look at Multi-Agent Debate Strategies for LLMs](https://mlsanity.com/paper/2311.17371/) (2023)
- [Let Me Speak Freely? A Study on the Impact of Format Restrictions on Performance of Large Language Models](https://mlsanity.com/paper/2408.02442/) (2024)
- [Single-Agent LLMs Outperform Multi-Agent Systems on Multi-Hop Reasoning Under Equal Thinking Token Budgets](https://mlsanity.com/paper/2604.02460/) (2026)
- [Can LLM Agents Really Debate? A Controlled Study of Multi-Agent Debate in Logical Reasoning](https://mlsanity.com/paper/2511.07784/) (2025)
- [Verbalized Confidence Triggers Self-Verification: Emergent Behavior Without Explicit Reasoning Supervision](https://mlsanity.com/paper/2506.03723/) (2025)
- [DeliberationBench: When Do More Voices Hurt? A Controlled Study of Multi-LLM Deliberation Protocols](https://mlsanity.com/paper/2601.08835/) (2025)

## Abstract (arXiv)

A common multi-agent design asks agents to report confidence and lets the highest-scoring agent speak next, implicitly using one scalar both to route the conversation and to estimate uncertainty. We audit this confidence-routed broadcast protocol by separating three trace-level questions: whether it selects the right candidate (routing), whether reported confidence behaves like a probability (calibration), and whether the selected agent publicly states the answer that won the turn (commitment). Our primary study covers 4,181 gpt-oss-120b olympiad-math traces; we repeat the audit on a 2-by-2 actor-by-benchmark grid that adds gemma-4-31B-it and a biology multiple-choice benchmark. In the primary cell, confidence discriminates correct from wrong candidates (AUROC 0.72) but is strongly overconfident (79% mean stated confidence versus 52% accuracy). A cross-fitted, tier-stratified isotonic procedure reduces Expected Calibration Error from 0.278 to 0.008 on held-out candidates, but it does not recover missing discrimination: raw AUROC is only 0.537 and 0.440 in the two Gemma cells. Routing is likewise setting-dependent. Fixed routers differ by at most 1.1 percentage points on gpt-oss/math, whereas raw-confidence argmax performs 5.6 and 11.2 points below random-valid selection in the Gemma cells. Commitment is distinct again: in the primary cell, poll and spoken answers diverge in 20.4% of valid pairs, 62.4% of those revisions are fresh generations, and the unconditional correctness shift is -1.7 points; the other three cells instead range from +0.9 to +12.2 points. The transferable lesson is procedural: routing discrimination, probability calibration, and public commitment must be measured separately before raw confidence is used for deployment decisions.