AI Paper Atlas

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

ZIP-RC: Zero-overhead Inference-time Prediction of Reward and Cost for Adaptive and Interpretable Generation

Rohin Manvi, Joey Hong, Tim Seyde, Maxime Labonne, Mathias Lechner, Sergey Levine

ICLR 2026 · Large language models › Reasoning and chain of thought · approach: Test-time compute scaling

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 ZIP-RC, a zero-overhead inference-time method that lets LLMs predict their own success probability (reward) and remaining computation cost (cost) to guide adaptive generation decisions.

Problem

LLMs lack introspection: they cannot anticipate their own likelihood of success or the compute needed, which hampers decisions on effort allocation, multiple attempts, stopping, and signaling success or failure.

Approach

Equips the model to predict reward and cost at inference time with essentially no extra overhead, enabling interpretable, adaptive meta-cognitive control over generation.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace majority voting with ZIP-RC sampling at matched costQwen3-1.7B · AIME 2024accuracy: 53.1 -> 65.8Improved
replace majority voting with ZIP-RC sampling at matched costLFM2-350M · AIME 2024accuracy: 26.9 -> 38.8Improved
replace majority voting with ZIP-RC sampling at matched costLFM2-1.2B · Mixed-difficultyaccuracy: 83.8 -> 86.0Improved
replace majority voting with ZIP-RC samplingQwen3-1.7B · Mixed-difficultyaccuracy: 91.0 -> 92.2Improved
replace sampling-utility optimisation with reward-threshold pruning (ablation)ZIP-RC · Mixed-difficulty (Qwen3-1.7B)accuracy: 88.9 -> 92.2Improved
replace sampling-utility optimisation with reward-threshold pruning (ablation)ZIP-RC · AIME 2024 (LFM2-1.2B)accuracy: 57.5 -> 60.9Improved
ZIP-RC sampling versus majority voting on costZIP-RC · Matched cost, LFM2-1.2B and LFM2-350Mgeneration cost: 1.60 -> 1.35; 1.70 -> 1.49Improved
ZIP-RC sampling versus majority voting on cost for Qwen3-1.7BZIP-RC · Matched costgeneration cost: 1.40 -> 1.43No clear effect
add length-based pruning to majority voting (ablation)Majority voting · AIME 2024 (Qwen3-1.7B)accuracy: 53.1 -> 25.1Hurt
keep full model trainable with KL regularisation instead of ZIP-RC-Lite head-only training (ablation)ZIP-RC · Qwen3-1.7B end-of-generation predictionaccuracy: 0.71 -> 0.88Improved
ZIP-RC-Lite in compute-bound setting (ablation)ZIP-RC-Lite · alpha=1.0compute efficiencyHurt

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