Resample or Reroute? Recoverable Stopping Debt Without Identified Action Selection
Not on the map. Why?
Abstract (arXiv)
After a weak verifier accepts a large-language-model response, a second call may resample or reroute. Because correctness is hidden, action selection is an identification problem. We order three gates: recoverable stopping debt, two-sided FIT action support, and held-out value from an outcome-blind selector. In a pinned 152-query MBPP+ experiment, a Qwen2.5-14B Base-only false-positive stop leaves +2.592 percentage points of Qwen2.5-7B recovery (query-cluster 95% interval [+1.618, +3.664]). Separately, after 7B Base-test rejection, fixed escalation to 14B exceeds leave-one-out 7B resampling by +2.882 points [+0.931, +5.201]; this is fixed-action ranking, not conditional selection. An all-episode audit produces a +2.697-point realized-maximum gap, but for two actions this statistic equals (1/2)E|Delta| - (1/2)|E Delta| and contains no observable-history term. It lies inside an exact-fold exchangeable reference (mean +3.158; 95% interval [+2.434, +3.947]). The audit unconditionally acts on 1,520 episodes: 1,240 observable stops and 280 verifier rejections; 198 stops are evaluator-only false positives. Neither tested outcome-blind controller improves on fixed rerouting. A separate LiveCodeBench ladder has all-zero FIT action advantages despite exclusive TEST rescues. A preregistered BigCodeBench support gate then finds only 23/19 and 22/19 signed episodes/queries against minima of 25/20, so L1-L4, DEV, and TEST stay unopened. Stopping debt exists, but current evidence does not identify when to resample rather than reroute.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 40 of its 54 references that are in the map. First posted on arXiv 9 Jul 2026.
Builds on
40 of 54 references.
References · 40
- 2024The Llama 3 Herd of Modelscited by 14,715
- 2021Training Verifiers to Solve Math Word Problemscited by 10,109
- 2021Evaluating Large Language Models Trained on Codecited by 8,320
- 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
- 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
- 2024Qwen2.5 Technical Reportcited by 4,942
- 2023Let's Verify Step by Stepcited by 3,789
- 2021Program Synthesis with Large Language Modelscited by 3,637
- 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,879
- 2024LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Codecited by 2,004
- 2023Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generationcited by 1,391
- 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,298
28 more
- 2023FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performancecited by 853
- 2024RouteLLM: Learning to Route LLMs with Preference Datacited by 682
- 2024BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructionscited by 546
- 2022CodeT: Code Generation with Generated Testscited by 384
- 2014On the Complexity of Best Arm Identification in Multi-Armed Bandit Modelscited by 286
- 2023LEVER: Learning to Verify Language-to-Code Generation with Executioncited by 239
- 2024RouterBench: A Benchmark for Multi-LLM Routing Systemcited by 238
- 2025Scalable Best-of-N Selection for Large Language Models via Self-Certaintycited by 187
- 2023AutoMix: Automatically Mixing Language Modelscited by 125
- 2023Let's Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with LLMscited by 115
- 2025NVIDIA Nemotron Nano 2: An Accurate and Efficient Hybrid Mamba-Transformer Reasoning Modelcited by 108
- 2024A Unified Approach to Routing and Cascading for LLMscited by 80
- 2025Router-R1: Teaching LLMs Multi-Round Routing and Aggregation via Reinforcement Learningcited by 72
- 2024Learning to Route LLMs with Confidence Tokenscited by 67
- 2025BEST-Route: Adaptive LLM Routing with Test-Time Optimal Computecited by 67
- 2025RouterEval: A Comprehensive Benchmark for Routing LLMs to Explore Model-level Scaling Up in LLMscited by 61
- 2024The Limits of Inference Scaling Through Resamplingcited by 53
- 2026LLMRouterBench: A Massive Benchmark and Unified Framework for LLM Routingcited by 38
- 2025Causal LLM Routing: End-to-End Regret Minimization from Observational Datacited by 28
- 2025Do We Truly Need So Many Samples? Multi-LLM Repeated Sampling Efficiently Scales Test-Time Computecited by 20
- 2025Too Consistent to Detect: A Study of Self-Consistent Errors in LLMscited by 18
- 2024Agreement-Based Cascading for Efficient Inferencecited by 16
- 2026LLMs Encode Their Failures: Predicting Success from Pre-Generation Activationscited by 14
- 2025Strategic Scaling of Test-Time Compute: A Bandit Learning Approachcited by 8
- 2025Semantic Agreement Enables Efficient Open-Ended LLM Cascadescited by 7
- 2024Model Cascading for Code: A Cascaded Black-Box Multi-Model Framework for Cost-Efficient Code Completion with Self-Testingcited by 5
- 2025Lookahead Routing for Large Language Modelscited by 5
- 2026What Does a Routing Oracle Measure Under Stochastic Decoding? Coupling, Scorer Choice, and Single-Commit Ceilingscited by 5
Built on by
Citing papers per year: 2026: 1.
1 paper.