AI Paper Atlas

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

CATS: Category-Aware Token-level Steering for Training-Free Redundancy Reduction in Large Reasoning Models

Mengfei Zhang, Zhenglin Wang

AAAI 2026 · Large language models › Reasoning and chain of thought · approach: Efficient and adaptive thinking · 0 citations (OpenAlex)

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 CATS, a training-free token-level steering method that categorizes redundancy in large reasoning model chain-of-thought by whether it harms the final answer, to reduce reasoning redundancy.

Problem

Large reasoning models produce excessive redundancy in chain-of-thought reasoning, which lowers inference efficiency and raises compute cost.

Approach

Taxonomizes LRM redundancy into destructive and non-destructive (benign) categories based on whether it affects the final answer, and applies category-aware token-level steering to suppress redundancy without training.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply CATS category-aware steeringDeepSeek-R1-Distill-Qwen-7B · MATH-500token length / accuracy: 3817.48 -> 2855.45 tokens; 92.96 -> 93.20Improved
apply CATS category-aware steeringDeepSeek-R1-Distill-Qwen-1.5B · MATH-500accuracy: 83.21 -> 82.01 with length 4528.31 -> 3577.36Mixed
apply CATS category-aware steeringDeepSeek-R1-Distill-Llama-8B · AIME 2024accuracy: 44.80 -> 46.14Improved
use six category vectors instead of one generic longest-minus-concise vector (ablation)CATS (Qwen-7B) · AIME 2024accuracy: 48.66 -> 52.32Improved
use six category vectors instead of one generic vector (ablation)CATS (Qwen-1.5B) · MATH-500accuracy: 79.55 -> 82.01Improved
apply generic intervention (single vector) (ablation)Qwen-1.5B · MATH-500accuracy: 83.21 -> 79.55Hurt
apply CATS to reduce logical driftDeepSeek-R1-Distill-Qwen-1.5B · Logical Drift occurrence rateoccurrence rate: 39.09 -> 31.27Improved
include the Logical Drift intervention vector (ablation)CATS · average over benchmarksaccuracy / length: accuracy -4.2% and length +2.0% when removedImproved
include the Repetition intervention vector (ablation)CATS · average over benchmarkslength / accuracy: length +4.3% when removed; accuracy change less than 1.0%Improved

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