Alignment, interpretability, red-teaming, bias and privacy: the research on what these systems do when they misbehave.
arXiv:2610.03176v1 Announce Type: new Abstract: We study hindsight-guided distillation for rare disease diagnosis on ZebraMap: a 1.5B student is fine-tuned on chain-of-thought traces from a 8B teache...
arXiv:2610.02549v1 Announce Type: new Abstract: Time and event expression extraction are fundamental temporal reasoning tasks, but the problem remains difficult due to annotation ambiguity, domain se...
arXiv:2610.02364v1 Announce Type: new Abstract: Explainability is increasingly required for perception models in intelligent vehicles, yet whether explanations remain faithful under driving domain sh...
arXiv:2610.03599v1 Announce Type: new Abstract: High-fidelity 3D head avatars have reached near-photorealistic quality. While recent methods enable text-driven manipulation, they struggle to provide...
arXiv:2610.03423v1 Announce Type: new Abstract: Reinforcement learning (RL) for Text-to-3D (T23D) generation requires optimization across multiple quality dimensions such as semantic alignment and te...
arXiv:2610.03252v1 Announce Type: new Abstract: Generative models of human-object interaction are bounded by the data that exists: everyday activities involve several objects, but most captured datas...
arXiv:2610.03142v1 Announce Type: new Abstract: Deep neural networks remain vulnerable to adversarial perturbations, which can distort not only predictions but also confidence scores, undermining unc...
arXiv:2606.20627v2 Announce Type: replace Abstract: Joint-Embedding Predictive Architectures (JEPAs) enable agents to plan in latent space by imagining the outcomes of candidate actions, yet task spe...
ULTRADISCOVERY is a benchmark that tests abductive exploration in an epistemically open, interconnected universe. It features five domains where agents must revise an initial theory and predict outcomes of unseen cross‑domain interventions, with a $2 imes 2$ design manipulating representation openness and evidence alignment. Experiments show that agents struggle to compose evidence into transferable representations, especially when the representation is open, and that even with aids exact predictions are rare without large budgets.
arXiv:2610.02959v1 Announce Type: new Abstract: Recent text-to-image models have made substantial progress in photorealism, aesthetics, and text-image alignment. Yet visually appealing images can sti...
The paper presents a mechanistic defense for Vision‑Language‑Action (VLA) models against adversarial patches. By using a sparse autoencoder, the authors identify a feature whose activation correlates strongly with the presence of an adversarial patch and suppress this feature only when a linear probe detects an attack. This conditional intervention improves robustness on the LIBERO‑10 benchmark while avoiding the performance degradation that occurs with continuous suppression.
The paper demonstrates that a targeted adversarial perturbation can reduce a vision‑language model’s training loss to near zero for a fixed target caption, yet the same model, when generating freely, still produces the correct description. This phenomenon, termed the train/inference gap, is traced to a single autoregressive step where the target token’s rank is fixed across all images, and further analysis shows that the language decoder, rather than the visual encoder, determines whether the corrupted signal is amplified or suppressed. The study uses a controlled two‑stage PGD attack on Qwen2.5‑VL‑7B‑Instruct and evaluates the effect on 200 held‑out COCO images, revealing that adversarial robustness in autoregressive VLMs largely depends on the language decoder’s prior. whyItMatters":"The findings suggest that defenses and faithfulness evaluations for deployed vision‑language models should focus on the language decoder rather than the visual encoder, as the former is the key determinant of robustness to adversarial perturbations."
SimplexUQ introduces the first benchmark and reproducible protocol for evaluating how conformal prediction wrappers allocate coverage across simplex‑valued predictions. The framework, called SimplexTasks‑12, combines six synthetic regimes and six real tasks (e.g., class probabilities, topic mixtures, spectral abundances) to compare existing wrappers on metrics such as marginal coverage, worst‑stratum coverage, max disparity, and computational cost. Empirical results show that no single wrapper consistently dominates, with Mondrian and BatchMVP performing best in different settings, and that removing predictor bias only partially mitigates disparity.
The paper introduces a new primal–dual algorithm for episodic adversarial linear constrained Markov decision processes (CMDPs) with unknown transitions. It achieves a rate‑optimal ×O(√K) regret and cumulative constraint violation, improving upon the previous ×O(K^{3/4}) bound and eliminating the need for Slater’s condition. The method combines adaptive FTRL, contracted value estimation, and an exponential Lyapunov function, enabling uniform concentration over the value function class and computational efficiency independent of the state‑space size.
HADRec is a Hierarchy-Aware Drug Recommendation framework that fuses molecular knowledge and electronic health records to improve medication recommendation. It uses LLaMA-7B to encode clinical notes, ChemBERTa to encode drug SMILES strings, and a cross‑attention mechanism for multimodal fusion, while a hierarchical predictor and consistency constraint loss enforce adherence to the ATC classification system. Experiments on MIMIC‑III and MIMIC‑IV show state‑of‑the‑art performance, strong generalization, and well‑calibrated predictions, with counterfactual evaluation indicating clinically aligned reasoning.
arXiv:2610. 01377v1 Announce Type: new Abstract: We study causal logistic bandits with counterfactual fairness constraints.
The paper examines how fairness conclusions in ICU mortality prediction using MIMIC-IV depend on the choice of metrics and the granularity of demographic analysis. It compares predictive-utility and subgroup-error metrics across various fairness interventions and introduces a lightweight adaptation strategy that balances ethnicity, gender, and insurance representation without conditioning on mortality outcomes. The study finds that different interventions can be evaluated differently across accuracy, sensitivity, and false-positive rate, and that marginal demographic summaries may hide heterogeneous error patterns within intersectional subgroups.
The paper introduces a method called Debias Anything that jointly addresses fairness and diversity in diffusion models without requiring sensitive-attribute annotations. By connecting a frozen diffusion model to a pretrained vision-language embedding space via an adapter, the approach uses pairs of text prompts to guide batch composition toward desired attribute proportions and employs a disagreement score to promote diversity. The method is applicable to both unconditional and text-conditional diffusion models and demonstrates improved quality and diversity while maintaining comparable fairness levels in experiments.
The paper investigates mechanistic interpretability, focusing on how automated circuit discovery is evaluated. It shows that the commonly used faithfulness objective can favor circuits that reproduce a model’s behavior poorly, creating an objective-level recovery gap. Experiments on four human-reference tasks and InterpBench reveal that many discovery methods misrank candidate circuits, and that restoring excluded signals can correct most of these misrankings without altering the circuits’ behavior.
AnchorPrompt is an adaptation technique for large audio‑language models that keeps the base model frozen and learns a single block of prompt vectors inserted at the decoder input. By training these prompts through self‑distillation on diverse audio and text perturbations, the method improves answer consistency and reduces hallucinations across multiple benchmarks. The approach is perturbation‑agnostic at inference, enabling zero‑shot transfer to unseen distortions such as reverberation and choice permutations.