Data-aware post-training quantization (PTQ) minimizes a per-token reconstruction loss on a small calibration corpus—an objective natural for average-token preservation, but one that implicitly weights positions by their empirical frequency. For **Automatic Speech Recognition (ASR)**, this weighting is misaligned with tail-sensitive recognition risk: names, numerals, technical terms, and domain-specific words are practically important but receive proportionally little calibration mass.

We propose **Tail-Aware Reconstruction Quantization (TARQ)**, a label-free PTQ framework that shifts calibration toward the lexical tail. **TARQ** introduces **rareBAL**, a closed-form per-Linear-layer rule that equalizes the calibration mass contributed by common and tail positions, plus a metric-consistent residual correction that aligns the sequential PTQ sweep with the reweighted objective. **TARQ** requires no entity labels, no curated calibration set, no validation decoding, and no additional training.

Across eight ASR backbones and eight datasets under 4-bit weight-only quantization, **TARQ** improves mean rare-WER, the **Word Error Rate (WER)** restricted to rare-word reference positions, while maintaining competitive aggregate WER. It achieves the lowest cross-corpus rare-WER swing, **0.63 percentage points (pp)**, among compared methods, and transfers without modification to entity-rich evaluation on ProfASR and ContextASR-Speech-En with no entity supervision.

Large language models (LLMs) require continual updates to rectify outdated or erroneous knowledge. Model editing has emerged as a compelling paradigm for introducing targeted modifications without the computational burden of full retraining. Existing approaches are mainly based on a locate-then-edit framework. However, in sequential editing contexts, where multiple updates are applied over time, they exhibit significant limitations and suffer from catastrophic interference, i.e., new edits compromise previously integrated updates and degrade preserved knowledge. To address these challenges, we introduce EvoEdit, a novel editing strategy that mitigates catastrophic interference through sequential null-space alignment, enabling stable and efficient model editing. By performing sequential null-space alignment for each incoming edit, EvoEdit preserves both original and previously modified knowledge representations and maintains output invariance on preserved knowledge even across long edit sequences, effectively mitigating interference. Evaluations on real-world sequential knowledge-editing benchmarks show that EvoEdit achieves better or comparable performance than prior state-of-the-art locate-then-edit techniques, with up to 3.53× speedup. Overall, these results underscore the necessity of developing more principled approaches for designing LLMs in dynamically evolving information settings, while providing a simple yet effective solution with strong theoretical guarantees.

Speculative Decoding (SD) accelerates autoregressive large language model (LLM) inference by decoupling generation and verification. While recent methods improve draft quality by tightly coupling the drafter with the target model, the verification mechanism itself remains largely unchanged, relying on strict token-level rejection sampling. In practice, modern LLMs frequently operate in low-margin regimes where the target model exhibits weak preference among top candidates. In such cases, rejecting plausible runner-up tokens yields negligible information gain while incurring substantial rollback cost, leading to a fundamental inefficiency in verification.We propose Margin-Aware Speculative Verification, a training-free and domain-agnostic verification strategy that adapts to the target model’s local decisiveness. Our method conditions verification on decision stability measured directly from the target logits and relaxes rejection only when strict verification provides minimal benefit. Importantly, the approach modifies only the verification rule and is fully compatible with existing target-coupled speculative decoding frameworks. Extensive experiments across model scales ranging from 8B to 235B demonstrate that our method delivers consistent and significant inference speedups over state-of-the-art baselines while preserving generation quality across diverse benchmarks. The code is available at https://github.com/5SSjw/MARS.

Reinforcement Learning from Human Feedback (RLHF) often suffers from sparse reward signals, making credit assignment to tokens challenging. We propose Shapley Credit Assignment Rewards (SCAR), a method that uses Shapley values from cooperative game theory to distribute sequence-level rewards across tokens or text spans according to their marginal contributions. SCAR provides dense rewards without requiring auxiliary critique models or fine-grained human annotations, while offering a principled game-theoretic foundation for credit assignment. We show theoretically that SCAR preserves the original optimal policy, and empirically demonstrate across sentiment control, summarization, and instruction tuning tasks that it converges faster and achieves higher final rewards than standard RLHF and attention-based dense reward baselines. To improve scalability on long sequences, we further introduce an Owen value–inspired adaptive text segmentation strategy. Our findings suggest that SCAR provides a more effective and theoretically sound method for credit assignment in RLHF, leading to more efficient alignment of LLMs.

Pediatric serious illness communication (SIC) is critically important, yet scalable communication training for clinicians remains limited. Compared with other dialogue simulation settings, pediatric SIC poses additional challenges, including multi-party interactions, response to parental distress, and strong dependence on feedback dynamics. Existing LLM-based simulators optimize generic dialogue quality rather than curriculum-contingent behavior required for effective SIC training. In collaboration with educators and pediatric clinicians, we introduce the first benchmark suite and simulation framework tailored to pediatric SIC training. Our benchmarks, PitfallBench and DialogueBench, evaluate simulators both at the turn-level and across full dialogues. We further propose SIC-Agents, a self-improving framework that generates a clinician-editable skill document to guide simulator behavior. Our experiments show that SIC-Agents outperforms static expert prompting. To support future research, we release our benchmarks and a training interface for parent simulation in pediatric SIC.

Medication recommendation predicts medications for patient visits, but existing methods still face two key challenges. At the model level, traditional drug recommendation methods only predict structured drug codes with limited evidence grounding, while LLM agents can use richer clinical context but may lack safety verification and traceability. At the task level, existing benchmarks often use broad medication categories, which ignore subgroup-level safety differences and can lead to risk overestimation. We introduce the first fine-grained medication recommendation setting based on fourth-level ATC code generation. We propose Safe Prescription Agent (SafeRx-Agent), a knowledge-grounded multi-agent framework that uses patient context, external clinical knowledge, and safety verification to recommend traceable medication sets. Experimental results on MIMIC-III and MIMIC-IV datasets show that SafeRx-Agent improves fine-grained medication prediction accuracy while controlling drug interactions, contraindications, and medication set size. Code accompanying this work is available at https://github.com/HWu3150/SafeRx-Agent/tree/main.

Large language models (LLMs) often struggle with context fidelity, producing inconsistent answers when responding to questions based on provided information. Existing approaches either rely on expensive supervised fine-tuning to generate evidence post-answer or train models to perform web searches without necessarily improving utilization of the given context. We propose CARE, a novel native retrieval-augmented reasoning framework that teaches LLMs to explicitly integrate in-context evidence within their reasoning process with the model’s own retrieval capabilities. Our method requires limited labeled evidence data while significantly enhancing both retrieval accuracy and answer generation performance through strategically retrieved in-context tokens in the reasoning chain. Extensive experiments on multiple real-world and counterfactual QA benchmarks demonstrate that our approach substantially outperforms supervised fine-tuning, traditional retrieval-augmented generation methods, and external retrieval solutions. This work represents a fundamental advancement in making LLMs more accurate, reliable, and efficient for knowledge-intensive tasks.