PRISM: Probability Reallocation with In-Span Masking for Knowledge-Sensitive Alignment

• URL: http://arxiv.org/abs/2604.01682v1
• Date: Thu, 02 Apr 2026 06:35:04 GMT
• Title: PRISM: Probability Reallocation with In-Span Masking for Knowledge-Sensitive Alignment
• Abstract summary: We study an augmented SFT setting in which training instances include coarse sentence-level factuality risk labels and inter-sentence dependency annotations.<n>We propose textbfPRISM, a differentiable risk-gated framework that modifies learning only at fact-critical positions.
• Score: 18.04153868078285
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Supervised fine-tuning (SFT) with token-level hard labels can amplify overconfident imitation of factually unsupported targets, causing hallucinations that propagate in multi-sentence generation. We study an augmented SFT setting in which training instances include coarse sentence-level factuality risk labels and inter-sentence dependency annotations, providing structured signals about where factual commitments are weakly supported. We propose \textbf{PRISM}, a differentiable risk-gated framework that modifies learning only at fact-critical positions. PRISM augments standard SFT with a lightweight, model-aware probability reallocation objective that penalizes high-confidence predictions on risky target tokens, with its scope controlled by span-level risk weights and model-aware gating. Experiments on hallucination-sensitive factual benchmarks and general evaluations show that PRISM improves factual aggregates across backbones while maintaining a competitive overall capability profile. Ablations further show that the auxiliary signal is most effective when used conservatively, and that knowledge masking and model-aware reallocation play complementary roles in balancing factual correction and capability preservation.

Related papers

• BALTO: Balanced Token-Level Policy Optimization for Hallucination Mitigation [51.22170603236523]
  We propose a Balanced Token-level Policy Optimization framework for hallucination mitigation.<n>BalTO extracts checkable factual claims, verifies them against the reference context, and projects claim-level judgments to token-level labels.<n>Experiments on ConFiQA, RAGTruth, and FinLLM-Eval show that BALTO achieves the highest faithfulness across all six model--benchmark settings.
  arXiv  Detail & Related papers  (2026-06-14T16:25:59Z)
• Beyond Surface Statistics: Robust Conformal Prediction for LLMs via Internal Representations [36.24681446361556]
  We propose a conformal framework for question answering that uses internal representations rather than output-facing statistics.<n>We show that our method achieves a better validity--efficiency trade-off than strong text-level baselines.<n>These results suggest that internal representations can provide more informative conformal scores when surface-level uncertainty is unstable under distribution shift.
  arXiv  Detail & Related papers  (2026-04-17T16:28:31Z)
• Less Approximates More: Harmonizing Performance and Confidence Faithfulness via Hybrid Post-Training for High-Stakes Tasks [52.81286869496811]
  Large language models are increasingly deployed in high-stakes tasks, where confident yet incorrect inferences may cause real-world harm.<n>We propose to jointly optimize unsupervised Reinforcement Learning from Internal Feedback with reasoning-trace-guided Reasoning Distillation.<n>We introduce HyTuning, a hybrid post-training framework that adaptively reweights RD and RLIF via a PRG-style metric.
  arXiv  Detail & Related papers  (2026-04-09T16:50:11Z)
• Safe RLHF Beyond Expectation: Stochastic Dominance for Universal Spectral Risk Control [12.053369001333058]
  We propose a risk-sensitive alignment framework that replaces expected cost constraints with First-Order Dominance (FSD) constraints.<n>We show that RAD improves harmlessness over baselines while remaining competitive in helpfulness, and exhibits greater robustness on out-of-distribution evaluations.
  arXiv  Detail & Related papers  (2026-03-11T16:24:20Z)
• On Robustness and Chain-of-Thought Consistency of RL-Finetuned VLMs [15.301640007799735]
  We show that simple, controlled textual perturbations--misleading captions or incorrect chain-of-thought (CoT) traces--cause substantial drops in robustness and confidence.<n>To better understand these vulnerabilities, we analyze RL fine-tuning dynamics and uncover an accuracy-faithfulness trade-off.
  arXiv  Detail & Related papers  (2026-02-13T01:12:00Z)
• Steering Vision-Language Pre-trained Models for Incremental Face Presentation Attack Detection [62.89126207012712]
  Face Presentation Attack Detection (PAD) demands incremental learning to combat spoofing tactics and domains.<n>Privacy regulations forbid retaining past data, necessitating rehearsal-free learning (RF-IL)
  arXiv  Detail & Related papers  (2025-12-22T04:30:11Z)
• Open-World Deepfake Attribution via Confidence-Aware Asymmetric Learning [78.92934995292113]
  We propose a Confidence-Aware Asymmetric Learning (CAL) framework, which balances confidence across known and novel forgery types.<n>CAL consistently outperforms previous methods, achieving new state-of-the-art performance on both known and novel forgery attribution.
  arXiv  Detail & Related papers  (2025-12-14T12:31:28Z)
• A Unified and Stable Risk Minimization Framework for Weakly Supervised Learning with Theoretical Guarantees [33.15955234458642]
  Weakly supervised learning has emerged as a practical alternative to fully supervised learning when complete and accurate labels are costly or infeasible to acquire.<n>We propose a principled, unified framework that bypasses such post-hoc adjustments by formulating a stable surrogate risk grounded in the structure of weakly supervised data.
  arXiv  Detail & Related papers  (2025-11-28T00:57:04Z)
• Data-Driven Calibration of Prediction Sets in Large Vision-Language Models Based on Inductive Conformal Prediction [0.0]
  We propose a model-agnostic uncertainty quantification method that integrates dynamic threshold calibration and cross-modal consistency verification.<n>We show that the framework achieves stable performance across varying calibration-to-test split ratios, underscoring its robustness for real-world deployment in healthcare, autonomous systems, and other safety-sensitive domains.<n>This work bridges the gap between theoretical reliability and practical applicability in multi-modal AI systems, offering a scalable solution for hallucination detection and uncertainty-aware decision-making.
  arXiv  Detail & Related papers  (2025-04-24T15:39:46Z)
• TRACE: TRansformer-based Attribution using Contrastive Embeddings in LLMs [50.259001311894295]
  We propose a novel TRansformer-based Attribution framework using Contrastive Embeddings called TRACE. We show that TRACE significantly improves the ability to attribute sources accurately, making it a valuable tool for enhancing the reliability and trustworthiness of large language models.
  arXiv  Detail & Related papers  (2024-07-06T07:19:30Z)
• When Does Confidence-Based Cascade Deferral Suffice? [69.28314307469381]
  Cascades are a classical strategy to enable inference cost to vary adaptively across samples. A deferral rule determines whether to invoke the next classifier in the sequence, or to terminate prediction. Despite being oblivious to the structure of the cascade, confidence-based deferral often works remarkably well in practice.
  arXiv  Detail & Related papers  (2023-07-06T04:13:57Z)

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.