When Self-Consistency Backfires: Majority Vote Hurts the Majority of Hard Science Problems for Small LLMs

• URL: http://arxiv.org/abs/2608.11403v2
• Date: Sat, 15 Aug 2026 22:30:00 GMT
• Title: When Self-Consistency Backfires: Majority Vote Hurts the Majority of Hard Science Problems for Small LLMs
• Abstract summary: Self-consistency via majority vote reduces per-problem accuracy on most GPQA Diamond problems for small instruction-tuned models.<n>This version reports that the most natural repair also fails, and separates three signal failures that v1 treated as one.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Self-consistency via majority vote reduces per-problem accuracy on most GPQA Diamond problems for small instruction-tuned models: 56.6% of problems for Qwen2.5-7B and 65.7% for Llama-3-8B. The obvious remedy is a verifier-free confidence gate. This version reports that the most natural repair also fails, and separates three signal failures that v1 treated as one. A token-entropy gate fails for a measurement reason: averaged over a chain of some 602 tokens, the statistic is a measurement of the prose rather than of confidence in the answer. On Qwen2.5-7B-Instruct-Turbo, 198 problems at 64 samples each, a sample whose answer contradicts its own problem's plurality still emits that answer at a median margin of 20.52 nats, with 75.7% above 10 nats. Both quantities were pre-registered and tested once on 69 problems no exploratory analysis had read; both passed. The unit is the whole result: pooled across the benchmark the margin separates correct from incorrect samples by +0.0604 on the fraction above 10 nats [+0.0183, +0.1017], excluding zero; per-problem and paired it does not, at -0.0168 [-0.0527, +0.0182], crossing zero. The claim is not that token log-probabilities carry no information, but that a signal with real across-question discrimination is close to useless for the within-question decision a router faces. The plurality-agreement gate's failure remains without a mechanism, and we report it as an open problem. These new claims rest on one model: a registered second-model replication was sampled and could not be evaluated, and we report that rejection rather than the result. We separately report that on hosted serverless inference at a small budget, three reasoning-native models could not be evaluated, for three separately measured reasons; all three are downloadable, so this bounds what a metered per-token API buys rather than what is knowable.

Related papers

• Marginal Fidelity Does Not Establish User Simulation in Demographic Synthetic Survey Panels: Response Contracts, Support Collapse and Conditioning Failure [51.736723807086385]
  Population-marginal agreement is evidence about an elicitation contract and an estimand obtainable without simulated respondents, not evidence of individual simulation.<n>On nine aligned model-battery pairs, a no-persona population-prevalence query averages 6.27 MAE versus 12.39 for committed panels and wins all nine comparisons.
  arXiv  Detail & Related papers  (2026-09-07T10:22:22Z)
• Clean Engineering, Unstable Measurement: A Preregistered Reliability Failure of Black-Box LLM Observers on Shared Endpoints [4.41537705949485]
  Language-model judges now gate training data, score generations, and drive leaderboards.<n>We audited that assumption in two preregistered campaigns with every threshold fixed in advance.<n>Across 52,988 audited request attempts, same-window repeat rankings agreed at Spearman 0.400 against a required 0.90.<n>We distill the evidence into a three-level snapshot-identity ladder, eight design rules, and a reporting checklist.
  arXiv  Detail & Related papers  (2026-09-03T17:59:43Z)
• Phantom Gains: Auditing Self-Improvement Against a Measured Null [10.74609430659738]
  We audit three rounds of rank-$32$ LoRA self-training on Qwen3-8B against a frozen control pushed through the identical pipeline.<n>We identify seven measurement failures, each of which inverts a reported finding when its control is absent.
  arXiv  Detail & Related papers  (2026-08-20T17:30:14Z)
• What Would Fix This RAG Failure? Auditing Counterfactual Response with Paired Evidence Interventions [0.0]
  We introduce Pair-ID, an offline audit that holds one query, retrieval state, and reader constant, then crosses two operations, adding missing support and deleting verified nonsupport, to measure a same-failure counterfactual response vector.<n>A complete funnel over 19,981 benchmark queries identifies 11,105 eligible Qwen failures, from which a prospectively fixed SHA-256 ordering selects 1,200 before generating any sampled response.<n>Results show that evidence sensitivity occurs at meaningful rates in the hash-selected eligible-failure sample, is only partially predictable from the observed failure, and is conditional on the reader.
  arXiv  Detail & Related papers  (2026-08-09T22:43:27Z)
• Accurate Ensembles, Fragile Narratives: Multi-Scale Stacking and a Fidelity Audit of LLM-Generated Explanations for Credit Risk [0.0]
  Credit scoring increasingly relies on models whose decision logic cannot be read off their parameters.<n>A common proposal closes that gap with a language model: compute feature attributions, hand them to an LLM, and let it write the rationale.<n>We build such a system end to end and test whether the second half of the promise holds.
  arXiv  Detail & Related papers  (2026-08-08T13:22:14Z)
• When Does Combining Language Models Help? A Co-Failure Ceiling on Routing, Voting, and Mixture-of-Agents Across 67 Frontier Models [0.0]
  Multi-model LLM systems are used to beat single-model accuracy.<n>We show that their gain is capped by a quantity the field rarely reports.
  arXiv  Detail & Related papers  (2026-06-25T17:06:06Z)
• Reasoning Quality Emerges Early: Data Curation for Reasoning Models [58.56882815783977]
  We show that diverse and challenging reasoning examples can be identified using only the initial reasoning tokens.<n>Our method outperforms existing baselines by up to 1.7% while being 91% more token efficient.
  arXiv  Detail & Related papers  (2026-06-25T09:32:58Z)
• OQMD: Single-Qubit Rotation Control Improves Low-CNOT Multiclass Quantum Classification [4.669017653337737]
  This work studies Optimal Quantum Measurement Decoding (OQMD)<n>OQMD: optimizing how quantum outcomes are mapped to classical labels by training a readout layer before measurement.<n>Experiments use trainable triple single-qubit rotations as one concrete, hardware-native realization of OQMD.
  arXiv  Detail & Related papers  (2026-06-12T04:05:48Z)
• When to Think Deeply: Inhibitory Deliberation for LLM Reasoning [5.19759149737193]
  We propose IDPR, a framework for response-conditioned inhibitory deliberation.<n>IDPR first generates a concise intuitive answer and then uses an inhibition controller to decide.<n>IDPR achieves the highest corrective precision, showing that response-conditioned inhibition better identifies fast answers.
  arXiv  Detail & Related papers  (2026-06-04T21:57:27Z)
• Off-the-Shelf LLMs as Process Scorers: Training-Free Alternative to PRMs for Mathematical Reasoning [51.88950852117154]
  Chunk-Level Guided Generation is a training-free alternative that uses an off-the-shelf large language model as a process scorer.<n>We show that scoring variable-length reasoning steps with large-model likelihoods is unreliable due to a systematic length bias.<n>Chunk-Level Guided Generation produces substantially shorter reasoning traces than PRM guided search.
  arXiv  Detail & Related papers  (2026-06-01T04:43:36Z)
• Step-wise Rubric Rewards for LLM Reasoning [72.17879367869503]
  Reinforcement Learning with Verifiable Rewards (RLVR) is widely used to improve reasoning in large language models.<n>We find 18.2% of steps in correct responses are wrong yet positively rewarded.<n>Across six mathematical reasoning benchmarks, SRaR improves average accuracy over RaR by 3.57 points.
  arXiv  Detail & Related papers  (2026-05-17T07:08:14Z)
• Distributional Energy-Based Models for Uncertainty-Aware Structured LLM Reasoning [40.342912574072024]
  Large Language Models produce structured outputs such as travel plans or code solutions.<n>Individual reasoning steps may appear correct while the output as a whole violates budgets, fails test cases, or contradicts earlier deductions.<n>We propose a learned quality scorer with deterministic analytical constraint penalties for verifying structured LLM outputs.
  arXiv  Detail & Related papers  (2026-05-15T17:08:27Z)
• Large Language Models Decide Early and Explain Later [47.20810161393936]
  We study the evolution of predicted answers over reasoning steps using forced answer completion.<n>We show that simples, including probe-based stopping, can reduce reasoning token usage by 500 tokens per query.
  arXiv  Detail & Related papers  (2026-04-24T06:26:24Z)
• Benchmarking Uncertainty Calibration in Large Language Model Long-Form Question Answering [7.1559850008795385]
  Large Language Models (LLMs) are commonly used in Question Answering (QA) settings.<n>Existing UQ approaches remain weakly validated in scientific QA.<n>We introduce the first large-scale benchmark for evaluating UQ metrics in reasoning-demanding QA.
  arXiv  Detail & Related papers  (2026-01-30T20:02:34Z)
• Sample Smart, Not Hard: Correctness-First Decoding for Better Reasoning in LLMs [72.82403830490084]
  We argue that the decoding rule should be calibrated by correctness, not confidence alone.<n>We propose simple strategies that achieve this goal: Greedy-Threshold makes sampling greedy at very low confidence steps.<n>Together, our findings challenge prevailings about decoding under uncertainty and show gains across math and general reasoning benchmarks.
  arXiv  Detail & Related papers  (2025-10-07T14:46:12Z)
• Eigen-1: Adaptive Multi-Agent Refinement with Monitor-Based RAG for Scientific Reasoning [53.45095336430027]
  We develop a unified framework that combines implicit retrieval and structured collaboration.<n>On Humanity's Last Exam (HLE) Bio/Chem Gold, our framework achieves 48.3% accuracy.<n>Results on SuperGPQA and TRQA confirm robustness across domains.
  arXiv  Detail & Related papers  (2025-09-25T14:05:55Z)
• VeriThinker: Learning to Verify Makes Reasoning Model Efficient [52.74493506816969]
  Large Reasoning Models excel at complex tasks using Chain-of-Thought (CoT) reasoning.<n>Their tendency to overthinking leads to unnecessarily lengthy reasoning chains.<n>We introduce VeriThinker, a novel approach for CoT compression.
  arXiv  Detail & Related papers  (2025-05-23T14:17:56Z)
• Not All Votes Count! Programs as Verifiers Improve Self-Consistency of Language Models for Math Reasoning [24.386388107656334]
  We introduce Prove, a framework that leverages translated programs derived from natural language solutions as a verification mechanism.<n>Unlike vanilla majority voting, our approach filters out solutions whose corresponding program output is inconsistent with the generated solution, aggregating only those that pass verification.<n>Our results show that Prove consistently outperforms vanilla majority voting for solving mathematical reasoning tasks across all model sizes and datasets.
  arXiv  Detail & Related papers  (2024-10-16T14:24:55Z)
• PRover: Proof Generation for Interpretable Reasoning over Rules [81.40404921232192]
  We propose a transformer-based model that answers binary questions over rule-bases and generates the corresponding proofs. Our model learns to predict nodes and edges corresponding to proof graphs in an efficient constrained training paradigm. We conduct experiments on synthetic, hand-authored, and human-paraphrased rule-bases to show promising results for QA and proof generation.
  arXiv  Detail & Related papers  (2020-10-06T15:47:53Z)
• Selective Question Answering under Domain Shift [90.021577320085]
  Abstention policies based solely on the model's softmax probabilities fare poorly, since models are overconfident on out-of-domain inputs. We train a calibrator to identify inputs on which the QA model errs, and abstain when it predicts an error is likely. Our method answers 56% of questions while maintaining 80% accuracy; in contrast, directly using the model's probabilities only answers 48% at 80% accuracy.
  arXiv  Detail & Related papers  (2020-06-16T19:13:21Z)
• TACRED Revisited: A Thorough Evaluation of the TACRED Relation Extraction Task [80.38130122127882]
  TACRED is one of the largest, most widely used crowdsourced datasets in Relation Extraction (RE) In this paper, we investigate the questions: Have we reached a performance ceiling or is there still room for improvement? We find that label errors account for 8% absolute F1 test error, and that more than 50% of the examples need to be relabeled.
  arXiv  Detail & Related papers  (2020-04-30T15:07:37Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.