AI Paper Atlas

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

Adaptive Thinking: Large Language Models Know When to Think in Latent Space

Pingzhi Li, Bairu Hou, Yun Zhu, Yihao Feng, Ke Ye, Tao Lei, Zhifeng Chen, Tianlong Chen, Xianzhi Du

ICLR 2026 · Large language models › Reasoning and chain of thought · approach: Efficient and adaptive thinking

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.

Studies adaptive allocation of thinking budget for LLMs, deciding when to think in latent space versus explicit chain-of-thought, based only on the abstract provided.

Problem

The relationship between LLM capability, query complexity, and optimal test-time thinking budget allocation is poorly understood for achieving compute-efficient reasoning.

Approach

Adaptive thinking that lets LLMs decide when to think in latent space rather than generating explicit intermediate chain-of-thought, allocating budget per query.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
gate thinking with adapter-predicted self-consistency (Sonata)Qwen3-8B · Average (5 benchmarks)accuracy: 74.1 -> 75.3Improved
gate thinking with SonataQwen3-8B · Average (5 benchmarks)thinking tokens: 9154 -> 7535Improved
gate thinking with SonataQwen3-8B · GSM8Kthinking tokens: 1994 -> 890Improved
gate thinking with SonataQwen3-32B · Average (5 benchmarks)accuracy: 78.0 -> 77.8No clear effect
gate thinking with SonataQwen3-8B · GPQA (out-of-domain)accuracy: 60.1 -> 62.0Improved
gate thinking with SonataGPT-OSS-120B-High · Averageaccuracy: 86.6 -> 85.4Hurt
use constant 4096 thinking budget instead of SonataQwen3-8B · AIME25accuracy: 60.0 -> 30.0Hurt
use Sonata instead of binary self-judge promptQwen3-8B · Average (5 benchmarks)accuracy: 72.7 -> 75.3Improved
use self-consistency instead of LM logits entropy as proxy (ablation)Sonata · Average (4 benchmarks)accuracy: 66.3 -> 79.6Improved
use 2-layer MLP instead of linear adapter (ablation)Sonata · Average (4 benchmarks)accuracy: 76.8 -> 79.6Improved
use 3-layer MLP instead of 2-layer (ablation)Sonata · Average (4 benchmarks)accuracy: 79.6 -> 79.3No clear effect
reduce calibration set from 1000 to 100 samples (ablation)Sonata · Average (4 benchmarks)accuracy: 79.6 -> 79.0No clear effect

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