Probing Routing-Conditional Calibration in Attention-Residual Transformers

• URL: http://arxiv.org/abs/2605.09850v1
• Date: Mon, 11 May 2026 01:06:25 GMT
• Title: Probing Routing-Conditional Calibration in Attention-Residual Transformers
• Abstract summary: Internal routing traces are often used to show confidence in predictions.<n>These traces do not provide stable routing-specific evidence for post-hoc calibration beyond confidence.<n>We show that routing profiles achieve comparable performance once a confidence-only capacity is included as a control.<n>This setting should not be read as internal-state calibration until matched-confidence, bandwidth, capacity, and permutation controls rule out common confounds.
• Score: 17.576884726604902
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Post-hoc calibration is usually evaluated as a function of logits or softmax confidence alone, even as routing-augmented architectures increasingly accompany predictions with sample-specific internal routing traces and pair them with claims of calibration-relevant uncertainty. We ask a basic question: do these traces provide stable routing-specific evidence for post-hoc calibration beyond confidence? We study this in Attention-Residual transformers (Kimi Team, 2026) through a matched-confidence diagnostic suite that stratifies examples by routing-derived state, compares subgroup gaps against within-bin routing-permutation nulls, and evaluates matched post-hoc probes differing only in their auxiliary feature. Across our completed AR runs, scalar routing summaries do not provide stable evidence of routing-conditional miscalibration: weighted gaps remain small or seed-sensitive, and only $1$ of $30$ within-bin permutation tests rejects the conditional-null at $α=0.05$ (only on one seed; not stable across seeds in that cell). AR-CondCal, a minimal $2$-D Nadaraya--Watson probe on confidence and routing-depth variance, lies within the seed-variance band of matched confidence-only and predictive-entropy controls and does not reliably improve worst-routing-tertile ECE; bandwidth-sensitivity checks (Scott multiples, CV-NLL, global-ECE oracle) do not change this. A full-vector MLP over $(c, H_1, \ldots, H_L)$ can appear to improve over a linear confidence baseline, but the apparent gain disappears once a capacity-matched confidence-only MLP is included as a control, and shuffled routing profiles achieve comparable performance. Apparent routing-aware calibration gains in this AR setting should not be read as internal-state calibration until matched-confidence, bandwidth, capacity, and permutation controls rule out common confounds.

Related papers

• SVR: Self-Verifying Refinement via Joint Verdict-Confidence Reinforcement Learning for Adaptive Test-Time Compute [62.3458279176813]
  Self-Verifying Refinement is an oracle-free multi-turn reinforcement learning framework.<n>It learns to use self-verification as a compute-control policy.<n>It achieves a macro-average accuracy of 0.563 with only 2.99 inference turns on average.
  arXiv  Detail & Related papers  (2026-07-30T16:20:58Z)
• One Human, $N$ Agents: Audit-Budget Allocation for LLM Agent Fleets under Miscalibrated, Correlated Confidence [41.99844472131922]
  A single human must audit $N$ LLM agents under a budget of $B ll N$ audits per round.<n>We model this as budgeted noisy inspection over a two-level Gaussian copula and locate the miscalibration threshold $*$ past which confidence-ranked auditing is emphworse than random.
  arXiv  Detail & Related papers  (2026-07-30T14:52:36Z)
• Lottery Tickets Are Not Deployment Tickets [4.230349220081444]
  We study whether an accuracy-matched lottery ticket or another sparse challenger can replace an incumbent dense model without reconfiguring downstream decision logic.<n>Across experiments, sparse candidates repeatedly recover dense-reference accuracy yet remain behaviorally different.<n>In small-gap settings with fixed-threshold policy diagnostics, lottery-ticket replacement changes 7% to 10% of accept-review decisions.
  arXiv  Detail & Related papers  (2026-07-29T15:25:19Z)
• Conformal Changepoint Localization and Root Cause Analysis with Corrupted Observations [55.76952683833966]
  This paper proposes weighted CONCH (W-CONCH) and weighted CROC (W-CROC) to reduce confidence set size when data may be corrupted.<n>Experiments on image-based and real-world changepoint and root-cause benchmarks show that uncertainty-based weighting substantially reduces confidence set size.
  arXiv  Detail & Related papers  (2026-07-29T05:16:59Z)
• Expectation Consistency Loss: Rethink Confidence Calibration under Covariate Shift [7.333172294727063]
  General confidence calibration methods assume training and test data are independent and identically distributed.<n>This paper proposes an unsupervised domain adaptation loss to calibrate confidence of the target domain.<n>We prove that computing ECL loss has the same sample complexity as Expected Error (ECE) and provide a theoretically grounded mini-batch trainable scheme for ECL loss.
  arXiv  Detail & Related papers  (2026-05-20T11:45:39Z)
• Discovery of Hidden Miscalibration Regimes [52.452902154360565]
  Models may be systematically overconfident on some kinds of inputs and underconfident on others.<n>We define the corresponding miscalibration field and propose a diagnostic framework for estimating it.<n>Our approach learns a calibration-aware representation of the input space and estimates signed local miscalibration by kernel smoothing in the learned geometry.
  arXiv  Detail & Related papers  (2026-05-13T13:07:50Z)
• Correction and Corruption: A Two-Rate View of Error Flow in LLM Protocols [51.56484100374058]
  We propose a paired-outcome measurement interface for auditing a single protocol step on exact-match tasks.<n>For each instance, the interface records a baseline correctness bit and a post-step correctness bit.<n>These rates predict accuracy changes and define a reusable empirical interface testable across seeds, mixtures, and pipelines.
  arXiv  Detail & Related papers  (2026-04-20T13:25:40Z)
• Direct Learning of Calibration-Aware Uncertainty for Neural PDE Surrogates [0.0]
  Cross-regularized uncertainty learns uncertainty parameters during training using routed through a held-out regularization split.<n>The predictor is optimized on the training split for fit, while low-dimensional uncertainty controls are optimized on the regularization split to reduce train-test mismatch.<n>We instantiate the approach in Fourier Neural Operators and evaluate on APEBench sweeps over observed fraction and training-set size.
  arXiv  Detail & Related papers  (2026-02-11T17:57:20Z)
• CoRefine: Confidence-Guided Self-Refinement for Adaptive Test-Time Compute [10.548368675645403]
  CoRefine is a confidence-guided self-refinement method that achieves competitive accuracy using a fraction of the tokens.<n>The controller consumes full-trace confidence to decide whether to halt, re-examine, or try a different approach.<n>We extend this to CoRefine-Tree, a hybrid sequential-parallel variant that adaptively balances exploration and exploitation.
  arXiv  Detail & Related papers  (2026-02-09T17:44:41Z)
• UAT-LITE: Inference-Time Uncertainty-Aware Attention for Pretrained Transformers [11.741258610945259]
  UAT-LITE is an inference-time framework that makes self-attention uncertainty-aware.<n>It reduces Expected Error by approximately 20% on average relative to a fine-tuned BERT-base baseline.<n>It improves selective prediction and robustness under distribution shift.
  arXiv  Detail & Related papers  (2026-02-03T00:51:26Z)
• Agentic Confidence Calibration [67.50096917021521]
  Holistic Trajectory (HTC) is a novel diagnostic framework for AI agents.<n>HTC consistently surpasses strong baselines in both calibration and discrimination.<n>HTC provides interpretability by revealing the signals behind failure.
  arXiv  Detail & Related papers  (2026-01-22T09:08:25Z)
• Unsupervised Conformal Inference: Bootstrapping and Alignment to Control LLM Uncertainty [49.19257648205146]
  We propose an unsupervised conformal inference framework for generation.<n>Our gates achieve close-to-nominal coverage and provide tighter, more stable thresholds than split UCP.<n>The result is a label-free, API-compatible gate for test-time filtering.
  arXiv  Detail & Related papers  (2025-09-26T23:40:47Z)
• Coverage-Guaranteed Speech Emotion Recognition via Calibrated Uncertainty-Adaptive Prediction Sets [0.0]
  Road rage, often triggered by emotional suppression and sudden outbursts, significantly threatens road safety by causing collisions and aggressive behavior.<n>Speech emotion recognition technologies can mitigate this risk by identifying negative emotions early and issuing timely alerts.<n>We propose a novel risk-controlled prediction framework providing statistically rigorous guarantees on prediction accuracy.
  arXiv  Detail & Related papers  (2025-03-24T12:26:28Z)
• Beyond calibration: estimating the grouping loss of modern neural networks [68.8204255655161]
  Proper scoring rule theory shows that given the calibration loss, the missing piece to characterize individual errors is the grouping loss. We show that modern neural network architectures in vision and NLP exhibit grouping loss, notably in distribution shifts settings.
  arXiv  Detail & Related papers  (2022-10-28T07:04:20Z)
• Distribution-free binary classification: prediction sets, confidence intervals and calibration [106.50279469344937]
  We study three notions of uncertainty quantification -- calibration, confidence intervals and prediction sets -- for binary classification in the distribution-free setting. We derive confidence intervals for binned probabilities for both fixed-width and uniform-mass binning. As a consequence of our 'tripod' theorems, these confidence intervals for binned probabilities lead to distribution-free calibration.
  arXiv  Detail & Related papers  (2020-06-18T14:17:29Z)

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.