arXiv AI
Aug 5
Enhancing VLM Reward Models Through Structure-Aware Fine-Tuning
arXiv:2608. 03875v1 Announce Type: cross Abstract: Designing effective reward functions remains a major bottleneck in Reinforcement Learning (RL).
AI safety and alignment
Alignment, interpretability, red-teaming, bias and privacy: the research on what these systems do when they misbehave.
arXiv AI
Aug 5
An empirical evaluation of the risks of AI model updates using clinical data: stability, arbitrariness, and fairness
arXiv:2604. 23954v2 Announce Type: replace Abstract: Artificial Intelligence (AI) and Machine Learning (ML) models used in clinical settings are increasingly deployed to support clinical decision-making.
arXiv AI
Aug 5
Self-Guided Adaptive Safety Alignment: Synthesizing and Internalizing Guidelines in Reasoning Models
arXiv:2511. 21214v4 Announce Type: replace-cross Abstract: Explicit safety policies can improve reasoning-model safety, but their effective coverage may lag behind evolving jailbreak strategies.
arXiv AI
Aug 5
Which Modality Decides? Counterfactual Modality Attribution for Multimodal LLMs
arXiv:2608. 00076v2 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) increasingly support high-stakes decision making by combining complementary information from images and text.
arXiv Machine Learning
Aug 5
Learning to Translate from Soft to Hard LLM Prompts
arXiv:2605. 27642v2 Announce Type: replace-cross Abstract: Soft prompting, also known as continuous prompting, is a parameter-efficient method for tuning LLMs to specific tasks.
arXiv Machine Learning
Aug 5
Token Buncher: Shielding LLMs from Harmful Reinforcement Learning Fine-Tuning
arXiv:2508. 20697v4 Announce Type: replace Abstract: As large language models (LLMs) continue to grow in capability, so do the risks of harmful misuse through fine-tuning.
arXiv AI
Aug 5
Knowing the Form, Not the Function: Automatically Auditing Answer--Authority Decoupling in Legal Benchmarks
arXiv:2608. 02621v1 Announce Type: cross Abstract: Legal benchmarks typically score final answers even when models also state legal authority.
arXiv AI
Aug 5
CARE-Bench: Benchmarking Patient-Facing LLM Triage
arXiv:2608. 03731v1 Announce Type: new Abstract: Patient-facing medical LLMs and agents increasingly answer symptom questions before clinician contact, where the key safety question is what action the user should take next.
arXiv AI
Aug 5
The Transformer Revolution, Part 1: Dynamic Processing through Output- Weight Interconnections
arXiv:2608. 03921v1 Announce Type: new Abstract: This paper offers a new interpretation of the Transformer during inference.
arXiv AI
Aug 5
PASE: Leveraging the Phonological Prior of WavLM for Low-Hallucination Generative Speech Enhancement
arXiv:2511. 13300v1 Announce Type: cross Abstract: Generative models have shown remarkable performance in speech enhancement (SE), achieving superior perceptual quality over traditional discriminative approaches.
arXiv AI
Aug 5
$S^3$: Improving Agent Safety through Multi-Stage Defense
arXiv:2608. 02683v1 Announce Type: cross Abstract: Large Language Model (LLM) agents rely on multi-stage agentic workflows, with stages such as memory, planning, and tool execution, to accomplish complex tasks.
arXiv AI
Aug 5
LogitScope: A Framework for Analyzing LLM Uncertainty Through Information Metrics
arXiv:2603. 24929v2 Announce Type: replace Abstract: Understanding and quantifying uncertainty in large language model (LLM) outputs is critical for reliable deployment.
arXiv Machine Learning
Aug 5
Prescribed-Basis Coefficient-to-Coefficient Neural Operator for Partial Differential Equations
arXiv:2510. 10350v3 Announce Type: replace-cross Abstract: Operator learning provides a data-driven approach to approximating solution operators of partial differential equations, but its effectiveness depends strongly on how input and output functions are represented.
arXiv AI
Aug 5
When Oracle Conditioning Misleads Deployment: Conditioning-Availability Bias in Echocardiographic Segmentation
arXiv:2608. 03342v1 Announce Type: cross Abstract: Conditional segmentation models may be trained and evaluated with auxiliary signals cleaner than those available at deployment.
arXiv AI
Aug 5
Pin Once, Swap Light: Subspace-Aligned Centroid-Residual Training for Efficient Ultra-LoRA Serving
arXiv:2608. 03579v1 Announce Type: cross Abstract: Modern multi-tenant Low-Rank Adapters (LoRAs) serving systems concurrently host tens to hundreds of LoRA adapters.
arXiv AI
Aug 5
CollaFuse: Collaborative Diffusion Models
arXiv:2406. 14429v4 Announce Type: replace-cross Abstract: In the landscape of generative artificial intelligence, diffusion-based models have emerged as a promising method for generating synthetic images.
arXiv Machine Learning
Aug 5
UNVaMP: Neural Knowledge Tracing with Variational Regularization of Latent Knowledge Dynamics
arXiv:2608. 03811v1 Announce Type: new Abstract: We introduce the Unified Neural Variational Measurement of Proficiency (UNVaMP) architecture, a knowledge tracing method that integrates observed student-item interactions with internal memory to produce evolving latent representations of student knowledge.
arXiv AI
Aug 5
Evaluating Counterfactual Sensitivity to Patient Information in Medication-Safety Reasoning
arXiv:2608. 03028v1 Announce Type: new Abstract: Applying a valid medication-safety rule when its patient-specific conditions are not met can produce an incorrect decision.
arXiv AI
Aug 5
Long-term Traffic Scene Prediction via Polynomial Representations in Autonomous Driving
arXiv:2608. 03330v1 Announce Type: new Abstract: This thesis addresses fundamental challenges in traffic scene prediction for autonomous driving by introducing robust and computationally efficient models based on polynomial representations.
arXiv AI
Aug 5
MissClick: Exploiting Digit-Serialized Coordinates to Attack GUI Grounding Models
arXiv:2608. 03740v1 Announce Type: new Abstract: Recent GUI visual grounding models generate screen coordinates as sequences of digit tokens that are parsed into numerical values and mapped to executable clicks.