Alignment, interpretability, red-teaming, bias and privacy: the research on what these systems do when they misbehave.
Large language models (LLMs) frequently produce confident yet factually incorrect responses when user inputs contain misleading premises, a phenomenon we attribute to fact perturbations in the input....
Accurate regional near-surface temperature forecasting is fundamental to short-range weather services and downstream risk assessment. Existing deep learning-based regional forecasters commonly produce...
Hosted retrieval-augmented generation (RAG) and semantic search allow users to query valuable provider-held corpora, raising two competing demands: to hide each query and chosen result, yet reveal onl...
Safety evaluation is critical for assessing whether aligned Large Language Models (LLMs) remain robust against jailbreak attacks. Existing automated testing methods, however, largely rely on response-...
In 2026, we held the fourth iteration of the SHROOM Shared Task series: SHROOM-Visions (\textbf{S}hared-task on \textbf{H}allucinations and \textbf{R}elated \textbf{O}bservable \textbf{O}vergeneration...
Continuous sign language recognition (CSLR) aims to recognize gloss sequences from unsegmented sign videos under weak sequence-level supervision. However, existing methods rely on sentence-level gloss...
We introduce VGA-BenchV2, an extended human-aligned benchmark and optimization framework for jointly evaluating and improving video generation quality and aesthetic value. Built upon VGA-Bench, VGA-Be...
The paper introduces a dual‑dimensional framework called Automated Item Similarity Analysis (AISA) that uses Large Language Models to assess incidental content similarity in large‑scale assessments. It combines Structured Decomposition and Semantic Relatedness to capture both structural and semantic nuances that traditional metrics miss. Psychometric validation shows that LLM‑derived metrics better align with construct‑irrelevant local dependence and produce more coherent item groupings, and simulations in Computerized Adaptive Testing demonstrate improved estimation stability and reduced bias with minimal efficiency loss.
AgentWorld is a simulation framework that evaluates agentic information retrieval by incorporating diverse user personalities based on the Big Five (OCEAN) traits, stateful tool-use environments, and a pass$^k$ consistency metric with structured fault classification and partial-credit scoring. It includes a risk analyzer that uses Monte‑Carlo rollouts and advanced scoring methods to quantify trajectory brittleness and attack attribution. Experiments with conversational analytics, customer‑support agents, and adversarial stress‑testing demonstrate that personality variation reveals failure modes hidden by uniform testing, such as cross‑domain leakage, contextual drift, and significant quality gaps across personas.
The paper introduces AnTrap, a benchmark that injects dynamic perturbations into Android GUI agent execution to evaluate robustness against runtime anomalies. It presents a taxonomy of anomalies across four layers—State, Thinking, Action, and Round—with ten subcategories, and a pipeline that maintains task solvability while adding realistic adversarial conditions. Experiments on 16 leading GUI models show universal vulnerability, and reinforcement learning can mitigate some traps but not deep contextual ones like state deadlock.
Luce is a 3D representation that unifies geometry and physically based rendering (PBR) materials within a voxelized multimodal Gaussian cloud, using dedicated Gaussian primitives for each modality. A variational autoencoder compresses this representation into a unified material‑aware latent space, which a rectified‑flow transformer generates from a single image conditioned on multi‑layer features from a pretrained image encoder. The latent decodes into relightable PBR Gaussians and an optional textured mesh with a tangent‑space normal map, achieving state‑of‑the‑art single‑image‑to‑3D generation on Toys4K and improving CLIP image‑alignment scores on a benchmark of AI‑generated images.
MetaRAG introduces a belief-action aligned policy optimization framework for agentic retrieval-augmented generation (RAG). It incorporates Verify-first Action Generation and Internal Belief Probing to assess whether the current evidence is sufficient before taking an action, and uses a consistency reward gated by answer correctness to guide training. Experiments on seven public QA benchmarks demonstrate that MetaRAG improves the accuracy-efficiency trade-off over existing RL-based agentic RAG baselines, with benefits that transfer across research settings, optimizers, and model backbones.
The paper introduces Constraint‑Guided Enterprise Data Mapping (CGM), a neuro‑symbolic approach that uses schema‑grounded admissibility constraints to steer large language models (LLMs) in aligning enterprise data. CGM operates in three stages: defining constraints with metadata, generating candidates under relaxed constraints to ensure feasibility, and ranking them with a bounded LLM. Experiments show that hard constraints dramatically reduce candidate space and improve F1 scores, enabling small models to match or surpass large LLMs at a fraction of the cost while reducing expert effort.
The paper surveys the evolution of smart glasses from simple capture devices to first‑person intelligence platforms that integrate human perception, context, and action. It introduces a unified framework that formalizes data flow, hardware capabilities, and seven foundational capabilities, and presents an L0‑L5 hierarchy for capture to embodied action. The study also maps nine application scenes, proposes a nine‑dimensional deployment framework, and outlines an evidence ladder for evaluation and trustworthiness.
PlaceSeek is a human‑centered geospatial retrieval framework that maps natural‑language queries to street‑view images by decomposing queries into functional and affective sub‑intents. It uses a Semantic Grounding Module to verify that candidate images contain the physical evidence needed for the intended activity, and an Affective Alignment Module to re‑rank these candidates based on human urban perception judgments. Evaluated on 31,956 Milan street‑view locations, PlaceSeek achieves high precision and ranking metrics, outperforming several vision‑language baselines and demonstrating the importance of both physical grounding and affective alignment for complex urban spatial queries.
NeuronGuard is a fine‑tuning defense for large language models that hardens them against both jailbreak and neuron‑level attacks. It redistributes safety signals across many neurons by identifying safety‑critical ones with per‑layer linear classifiers, enforcing refusal behavior when those neurons are ablated, and applying KL‑divergence regularization for consistency. A randomized gradient projection preserves task performance, and the authors provide a formal guarantee that NeuronGuard lowers the attack success rate upper bound, with experiments showing near‑zero success rates across multiple models and attack strategies.
arXiv:2511.02605v3 Announce Type: replace Abstract: Shielding is widely used to enforce safety in reinforcement learning (RL), ensuring that an agent's actions remain compliant with formal specificat...
The paper introduces Restoring without Forgetting (RwF), a continual learning framework for image restoration that handles multiple degradations sequentially without accessing prior data. RwF trains a lightweight adapter for each new degradation, uses an unsupervised routing mechanism to select the correct restoration path, and achieves significant PSNR gains over fine‑tuning on five benchmark degradation domains. The method also demonstrates strong transfer performance on eleven real‑degradation datasets with high routing accuracy.
PHMForge is an evaluation environment that tests large language model agents on industrial prognostics tasks using the Model Context Protocol (MCP). It provides 99 SME-authored scenarios across eight asset classes, 39 MCP-native tools wrapping published PHM algorithms, and deterministic evaluators that separate protocol fluency, reasoning, instrumentation, and tool use. The benchmark shows that the best agent configuration achieves 80.8% pass@1, with orchestration and tool sequencing errors as the main failure modes, and demonstrates the limitations of static retrieval for prognostic computation.
The paper introduces Skill Abstraction with Interpretable Latents (SAIL), a method that models human skill as a persistent, multi‑dimensional construct inferred from naturalistic behavior over time. SAIL produces a robust skill embedding that blends expert and novice bases, learns transferable subskills through counterfactual subskill swaps, and supports skill‑informed behavior prediction across various in‑domain contexts. Experiments on racing and baseball demonstrate that SAIL achieves strong predictive performance, improves behaviorally grounded disentanglement compared to baselines, and enhances downstream AI coaching outcomes.