Diagnosing Correctness Probes under Self-Judgement Confounding

• URL: http://arxiv.org/abs/2607.16799v1
• Date: Sat, 18 Jul 2026 12:45:49 GMT
• Title: Diagnosing Correctness Probes under Self-Judgement Confounding
• Abstract summary: Hidden-state readouts can predict whether language-model outputs are correct, but objective correctness (OC) usually agrees with the model's own self-judgement ( SJ)<n>We construct conflict cases in which OC and SJ predict opposite readout orderings.<n>We estimate factorial SJ- and OC-associated directions and evaluate their polarity across mathematical reasoning and factual recall.
• Score: 1.2691047660244335
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Hidden-state readouts can predict whether language-model outputs are correct, but objective correctness (OC) usually agrees with the model's own self-judgement (SJ), leaving the decoded signal semantically ambiguous. We construct conflict cases in which OC and SJ predict opposite readout orderings. On high-confidence disagreements, conventional correctness-labelled contrasts often rank incorrect/self-endorsed responses above correct/self-rejected responses, following SJ rather than OC. We estimate factorial SJ- and OC-associated directions and evaluate their polarity across mathematical reasoning and factual recall. Across four instruction-tuned models up to 14B parameters, the SJ-associated direction transfers above chance in both cross-domain directions for every model, whereas the OC-associated direction has a below-chance point estimate for the expected OC ordering in every corresponding condition. This transfer asymmetry develops across middle-to-late layers, persists under answer-likelihood, sequence-length, and null-direction controls, and extends to MMLU and binary TruthfulQA without target-domain direction fitting. Across the studied models and diagnostic subsets, the most reliably transferable component preserves SJ-associated polarity. Transferability alone therefore does not establish objective-correctness semantics.

Related papers

• Symmetry-Aware Likelihood-Orbit Aggregation for Selective Left-Right Claim Verification [14.108900142001628]
  We introduce Relation-Orbit, a closed-form contrast that assigns eight normalized likelihoods to query-supporting and counterfactual roles.<n>On VSR and GQA across four frozen vision-language models, Relation-Orbit yields higher mean test coverage at a 10% selective-risk calibration target.
  arXiv  Detail & Related papers  (2026-09-15T11:19:03Z)
• CARD: Calibration via Agreement in Reverse Diffusion for Out-of-Domain MRI Segmentation [14.844162865265673]
  We propose CARD, which maps the temporal aggregate of disagreement to a temperature field applied per pixel across all classes, so that confidence changes while the segmentation does not.<n>Across cardiac, prostate and brain MRI shifts, CARD lowers calibration error in 45 of 49 comparisons against the strongest baseline in each setting.
  arXiv  Detail & Related papers  (2026-08-26T07:23:08Z)
• Why Does the Future Branch? Identifiable Closure Tests for Stochastic Physical World Models [0.0]
  ClosurePairs crosses compatible microstates with repeated disturbances.<n>It recovers source attribution at unchanged likelihood.<n>It reduces equal-budget decomposition error in a nonlinear interaction benchmark.
  arXiv  Detail & Related papers  (2026-08-01T11:07:24Z)
• Auditing Question-Order Effects in Large Language Models with the QQ Equality: Mechanism Characterization and a Saturation Caveat [4.07636450847048]
  Human survey respondents exhibit question-order effects that satisfy the QQ (quantum question) equality.<n>We develop the QQ equality into an audit criterion for sequential judgments of autoregressive large language models.
  arXiv  Detail & Related papers  (2026-07-19T12:22:06Z)
• Purified OPSD: On-Policy Self-Distillation Without Losing How to Think [52.879387744920415]
  We find that on-policy self-distillation (OPSD) consistently fails on long chain-of-thought (long-CoT) reasoning models.<n>We propose a two-step solution: first, we construct a reference-only teacher to isolate the non-transferable component of the supervision signal.<n>Second, we use pointwise mutual information (PMI) as the mechanism to transform this residual into a well-formed PMI target distribution.
  arXiv  Detail & Related papers  (2026-07-02T14:33:07Z)
• Dual-Branch Cross-Projection Debiasing through Diffusion-based Disentanglement [66.67463557274358]
  Foundation models trained on biased datasets often rely on spurious correlations between target labels and non-causal attributes.<n>We introduce Confidence-guided Concept Mining (CBCM) to identify reliable spurious attributes without attribute annotations.<n>We propose Dual-branch Cross-projection Debiasing (DCD), a prompt-tuning framework that separates target and spurious representations into two branches.
  arXiv  Detail & Related papers  (2026-06-23T05:28:47Z)
• Decoding in Order-Agnostic Language Models: Chain-Rule Deviation and Uniform Spreading [0.6916773850242582]
  Order-agnostic language models (OALMs) are trained to predict masked tokens under arbitrary conditioning sets.<n>We show that learning conditionals are not exact factorizations of a coherent joint distribution.<n>We propose a complementary diagnostic based on the shape of the confidence trace.
  arXiv  Detail & Related papers  (2026-05-31T04:25:36Z)
• Bias Leaves a Gradient Trail: Label-Free Bias Identification via Gradient Probes on Concept Decompositions [8.894428198928333]
  We present a bias-label-free, post-hoc method for identifying spurious concepts in frozen vision models.<n>On Colored MNIST and Waterbirds the method recovers concepts aligned with the known spurious cue.<n> suppressing the top-ranked concepts at inference time improves worst-group accuracy by up to 17.9 percentage points.
  arXiv  Detail & Related papers  (2026-05-27T17:39:10Z)
• When Individually Calibrated Models Become Collectively Miscalibrated [3.556355987197792]
  We show that individually calibrated predictors can become collectively miscalibrated when their predictions interact strategically.<n>We prove that under Brier-score-based aggregation with positively correlated beliefs, each agent's individually optimal report systematically underestimates the positive-class probability.<n>In contrast, VCG-based aggregation aligns incentives by rewarding marginal contribution, achieving dominant-strategy incentive compatibility and near-optimal performance.
  arXiv  Detail & Related papers  (2026-05-14T05:25:16Z)
• FACT-E: Causality-Inspired Evaluation for Trustworthy Chain-of-Thought Reasoning [49.65751420291115]
  Chain-of-Thought (CoT) prompting has improved LLM reasoning, but models often generate explanations that appear coherent while containing unfaithful intermediate steps.<n>We propose FACT-E, a causality-inspired framework for evaluating CoT quality.<n>We show that FACT-E improves reasoning-trajectory selection and yields stronger in-context learnings.
  arXiv  Detail & Related papers  (2026-04-12T15:35:08Z)
• Learning from Emptiness: De-biasing Listwise Rerankers with Content-Agnostic Probability Calibration [76.08899010904652]
  CapCal is a training-free framework that mechanically decouples positional bias from ranking decisions.<n>It achieves superior performance among training-free methods while preserving single-pass efficiency.
  arXiv  Detail & Related papers  (2026-04-11T10:47:22Z)
• CARE What Fails: Contrastive Anchored-REflection for Verifiable Multimodal [84.71254539482369]
  Group-relative reinforcement learning with verifiable rewards (RLVR) often wastes the most informative data it already has the failures.<n>We present CARE, a failure-centric post-training framework for multimodal reasoning that turns errors into supervision.<n> CARE improves accuracy and training smoothness while explicitly increasing the share of learning signal that comes from failures.
  arXiv  Detail & Related papers  (2025-12-22T16:34:21Z)
• 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)
• Judging with Confidence: Calibrating Autoraters to Preference Distributions [56.17041629492863]
  We argue that a reliable autorater must learn to model the full distribution of preferences defined by a target population.<n>We present two learning methods tailored to different data conditions.<n>Our results show that finetuning autoraters with a distribution-matching objective leads to verbalized probability predictions that are better aligned with the target preference distribution.
  arXiv  Detail & Related papers  (2025-09-30T20:36:41Z)

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.