# Does the Same Token Mean the Same State? MoE Routing as Signal for Reasoning Control

arXiv: https://arxiv.org/abs/2606.22798 · HTML: https://mlsanity.com/paper/2606.22798/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 14 of its 26 references that are in the map.
- First posted on arXiv 22 Jun 2026.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)
- Mixture-of-experts / [Expert routing strategies](https://mlsanity.com/moe/expert-routing-strategies/)

## Builds on (most cited first, 14 of 14)

- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](https://mlsanity.com/paper/2405.15793/) (2024)
- [GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding](https://mlsanity.com/paper/2006.16668/) (2020)
- [Reasoning Models Know When They're Right: Probing Hidden States for Self-Verification](https://mlsanity.com/paper/2504.05419/) (2025)
- [The Myth of Expert Specialization in MoEs: Why Routing Reflects Geometry, Not Necessarily Domain Expertise](https://mlsanity.com/paper/2604.09780/) (2026)
- [Model Utility Law: Evaluating LLMs beyond Performance through Mechanism Interpretable Metric](https://mlsanity.com/paper/2504.07440/) (2025)
- [Do LLMs Signal When They're Right? Evidence from Neuron Agreement](https://mlsanity.com/paper/2510.26277/) (2025)
- [Beyond Benchmarks: Understanding Mixture-of-Experts Models through Internal Mechanisms](https://mlsanity.com/paper/2509.23933/) (2025)
- [NEX: Neuron Explore-Exploit Scoring for Label-Free Chain-of-Thought Selection and Model Ranking](https://mlsanity.com/paper/2602.05805/) (2026)
- [In harmony with gpt-oss](https://mlsanity.com/paper/2604.00362/) (2026)
- [Layer-wise MoE Routing Locality under Shared-Prefix Code Generation: Token-Identity Decomposition and Compile-Equivalent Fork Redundancy](https://mlsanity.com/paper/2604.17182/) (2026)

## Built on by (most cited first, 2 of 2)

- [Disagree to Explore, Agree to Commit: Routing-Guided Test-Time Scaling for Software Agents](https://mlsanity.com/paper/2608.22191/) (2026)
- [From Concentration to Differentiation and Back: Routing Effective Rank in MoE Reasoning Cohorts](https://mlsanity.com/paper/2609.06403/) (2026)

## Abstract (arXiv)

In sparse Mixture-of-Experts language models, does the same token id imply the same router state and the same experts producing it? Holding the emitted token id fixed at repeated anchors, we find it does not: the experts that produce it still separate task context, trajectory history, and reasoning-effort mode. This residual structure supports test-time control: near boundary anchors (the final-response transition) and delimiter anchors (which open the answer, e.g. texttttextbackslash boxed or code fences), routing neighborhoods already align with final-answer basins at a marker-only readout and strongest when the routing is read at the answer opening. We operationalize this as RAD (Routing Agreement Decoding), an answer-string-free multi-rollout selector: it locates a fixed anchor, represents each rollout by its anchor-window MoE routing states, and returns the densest Weighted-Jaccard K-NN route-basin center, without parsing, normalizing, executing, or voting over answer strings. Across 10 sparse-MoE configurations (gpt-oss, Qwen3-MoE) and 6 datasets spanning math, GPQA, and code, RAD is on par with Majority where string voting is well-posed, with small positive paired deltas (RAD 73.9 / RAD+DC 74.2 vs. Majority 73.6). Like majority voting, RAD is not a verifier: a dense wrong basin can still win. Its value is the interface: the same selector gives direct pass@1 on code, where exact-string voting is ill-defined, and the same routing-density principle, re-anchored to the agentic boundary, improves best-of-16 patch selection on SWE-bench Verified over random, where patches have no answer string to vote on.