AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

Causal Dependency-Aware Unsupervised Routing for Large Reasoning Models

Jiacheng Liu, Hao Liu, Xiaofeng Hou, Wei Xue, Yike Guo

ICML 2026 · Large language models › Reasoning and chain of thought · approach: Routing, pruning and distillation for reasoning

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.

Proposes an unsupervised, causal dependency-aware router that selects among a pool of LLMs for reasoning queries without human-annotated preference data.

Problem

Routing queries to the most suitable LLM in a diverse pool normally requires expensive human-annotated preference data for supervised routers, which are brittle under distribution shift.

Approach

Unsupervised routing that models causal dependencies to select models without supervised preference labels; full text unavailable so details beyond the abstract are not confirmed.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use ReasoningRouter instead of RandomRouterrouting over reasoning-model pool · 9-benchmark average (BGE-M3)routing accuracy: 84.99 -> 88.20Improved
use ReasoningRouter instead of RandomRouterrouting over reasoning-model pool · 9-benchmark average (Embeddinggemma-300m)routing accuracy: 84.99 -> 88.23Improved
use thinking-and-answer scoring instead of monolithic OutputRouterReasoningRouter · 9-benchmark average (BGE-M3)routing accuracy: 87.81 -> 88.20Improved
use thinking-and-answer scoring instead of monolithic OutputRouterReasoningRouter · 9-benchmark average (Embeddinggemma-300m)routing accuracy: 88.16 -> 88.23No clear effect
route on thinking and answer jointly instead of thinking onlyReasoningRouter · TruthfulQA (BGE-M3)routing accuracy: 80.48 -> 81.98Improved
route on thinking and answer jointly instead of answer onlyReasoningRouter · TruthfulQA (BGE-M3)routing accuracy: 76.58 -> 81.98Improved
use ReasoningRouter instead of monolithic OutputRouterReasoningRouter · MathQA (Embeddinggemma-300m)routing accuracy: 93.27 -> 93.98Improved
vary proportion of thinking weight w_t versus answer weight w_a (ablation)ReasoningRouter · nine benchmarksrouting accuracyNo clear effect
increase the number of models in the candidate pool (ablation)ReasoningRouter · nine benchmarks (BGE-M3)average accuracyImproved

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