Truncated Reasoning Under a Budget: Evaluating System-1 Models as Decision-makers
We evaluate Jev as a decision reader of truncated reasoning across seven LLM reasoners and five multiple-choice benchmarks (35 pairs).
ProofPaper ↗
Key points
- Large language models (LLMs) often generate more reasoning tokens than needed.
- Limiting the reasoning budget reduces cost, but forcing an answer from an incomplete trace can hurt accuracy.
- System-1 decision models select among candidate options without producing a reasoning response.
- These findings suggest that System-1 decision models can complement LLM reasoning and reduce the need for costly reasoning completion.
Sources (1)
- [1]Truncated Reasoning Under a Budget: Evaluating System-1 Models as Decision-makersarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 04:29 AM
We evaluate Jev as a decision reader of truncated reasoning across seven LLM reasoners and five multiple-choice benchmarks (35 pairs).
Large language models (LLMs) often generate more reasoning tokens than needed.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 19, 2026PFedQFT: Personalized Federated Quantum-Inspired Fine-Tuning for Medical Large Language Models
- Oct 17, 2026A Cross-Domain Evaluation of Prompting Techniques for Large Language Models
- Oct 17, 2026Medical Fine-Tuning Reshapes, Not Reduces, Patient-Facing AI Risk: A Rubric-Based, LLM-Judge Evaluation Framework
- Oct 16, 2026"You" Is Not Toxic: How LLM Labelers Rewrite the Web's Conversation Laws
- Oct 16, 2026Characterising and Analysing Retrieval-Augmented Generation Techniques
- Oct 16, 2026NL2SQL with Iterative Self-Refinement