VideoDR is the first benchmark for open-domain video question answering, combining cross-frame visual extraction, iterative web retrieval, and multi-hop reasoning. Evaluation of multimodal models (closed/open-source) shows Agentic paradigm is not consistently superior to Workflow; key challenges are goal drift and long-horizon consistency.
Systematic investigation of diffusion models for end-to-end autonomous driving (E2E AD). Hyper Diffusion Planner (HDP) framework achieves 10x performance improvement over base model, tested on 200 km real-world driving and 6 urban scenarios. Includes reinforcement learning post-training to enhance safety and robustness.
Study on non-uniform replay effectiveness in off-policy RL. Authors identify three key factors: replay volume, recency of transitions, and entropy of sampling distribution. They propose Truncated Geometric replay, which biases toward recent experience while maintaining high entropy, improving sample efficiency in low-volume regimes.
Researchers propose UALM (Unified Agent Lifecycle Management), a governance framework for agentic AI systems in healthcare. The model addresses agent sprawl through five control-plane layers: identity registry, orchestration, PHI-bounded context, runtime policy enforcement with kill-switch triggers, and lifecycle management linked to credential revocation and audit logging.
Paper presents unsupervised denoising framework for low-dose liver CT using Cycle-GAN-inspired approach. Combines U-Net, attention mechanisms, residual networks with perceptual loss. Validated on real clinical dataset with radiologist evaluation.
SurgUn, an unlearning method for text-to-image diffusion models, treats forgetting as controlled competition rather than direct deletion. Using target-gradient ascent and descent over semantically diverse distractors, it reduces erase-retain imbalance and limits collateral damage. Tested on Stable Diffusion v1.5, SDXL, and SANA-1.5.
Complementary-label learning (CLL) methods remain limited to 10-class classification. This paper proposes BICL, a framework that deliberately uses biased (non-uniform) transition matrices to restrict complementary labels to class subsets. On CIFAR-100 and TinyImageNet-200, BICL achieves 7× accuracy improvements over traditional methods.
UxSID introduces a framework for modeling ultra-long user sequences using Semantic IDs and dual-level attention. The architecture balances computational efficiency with semantic awareness, achieving state-of-the-art performance and 0.337% revenue lift in large-scale advertising A/B testing.
Diagnostic study of catastrophic forgetting in continual learning using Sparse Autoencoders. Framework analyzes task-specific information evolution at latent concept level. Finding: most apparent concept-level forgetting is recoverable under linearity assumption; degradation stems from representational accessibility changes rather than complete information erasure.
GRAFT is a hybrid AFT model for survival analysis that decouples prognostic ranking from calibration of survival estimates. It combines a linear AFT model with a non-linear residual neural network and stochastic gates for feature selection. Trained on C-index-aligned ranking loss with conditional imputation, it outperforms baselines in discrimination and calibration.
Pilot benchmark for translating natural language to First-Order Logic (FOL) in planetary exploration. Dataset built from NASA mission documentation (2003-2013), manually annotated with FOL representations capturing temporal structure, agent roles, and operational dependencies. Structured predicate vocabularies provided.
SSL4RL leverages self-supervised learning tasks (image rotation, masked patch reconstruction) as reward signals for reinforcement learning fine-tuning of vision-language models. The framework eliminates the need for human preference data and improves performance on vision-centric and vision-language reasoning benchmarks.
MindMelody is a closed-loop real-time system generating personalized music from EEG signals. A Transformer-GNN model decodes EEG into emotional states (Valence-Arousal), an LLM with RAG formulates intervention plans, and a Hierarchical EEG Controller guides audio synthesis. A continuous feedback loop adapts parameters based on the user's evolving EEG dynamics.
R2M (Real-Time Aligned Reward Model) introduces a lightweight RLHF framework to mitigate reward overoptimization. Instead of relying solely on semantic representations, R2M leverages evolving hidden states from the policy model to align with real-time distribution shifts during reinforcement learning training.
Study showing that two protocols for evaluating layer redundancy in transformers (replacement and interchange) yield divergent results for identifying layers to prune. On Pythia, Qwen3-8B, and Llama-3.1-8B, the protocol gap dramatically changes which layers appear safe to remove, even under the same KL evaluator.
Asynchronous RL pipelines for LLM agents lose historical old logits required for PPO off-policy correction, entangling discrepancy repair with staleness correction. The paper proposes three acquisition strategies (snapshot, dedicated model, interruption) and a revised PPO-EWMA method to preserve decoupled correction semantics.
Controlled study comparing human vs synthetic soft-labels on MNIST. Human soft-labels improve model calibration and alignment with human uncertainty, beyond mere correction of mislabeled data. Shows primary value lies in regularization and stable convergence across training runs.
MarsTSC, an agentic framework for VLMs, addresses few-shot multimodal time series classification. Three collaborative roles (Generator, Reflector, Modifier) iteratively refine a knowledge bank via reflective reasoning. Tested on 12 benchmarks across 6 VLM backbones: consistent gains, interpretable rationales.
PCFJudge, an inference-time method, evaluates factuality by rerunning a listwise prompt across multiple candidate orderings and aggregating scores. On RewardBench 2 Factuality with K=7 permutations, top-1 accuracy improves from 86% to 91.33% (GPT-5.4) and 86.33% to 89.67% (Claude Sonnet 4.6).
Systematic empirical study of in-context learning capabilities in transformers on Gaussian-mixture binary classification tasks. Authors analyze how test accuracy depends on input dimension, number of in-context examples, and pre-training task diversity. They characterize benign overfitting emergence and identify critical parameter regions for ICL success.
SIPO stabilizes diffusion model alignment to human preferences by addressing training instability and off-policy bias. The method introduces DPO-C&M to clip uninformative timesteps and applies timestep-aware importance reweighting. Experiments on SD1.5, SDXL, CogVideoX-2B/5B, and Wan2.1-1.3B demonstrate improvements over Diffusion-DPO.
DoublyCal, a framework to improve LLM reliability by combining Knowledge Graphs and uncertainty calibration. A lightweight proxy model generates KG evidence with calibrated confidence, guiding a black-box LLM toward more accurate and well-calibrated predictions. Tested on knowledge-intensive benchmarks with reduced token costs.
Template Infilling (TI) is a conditioning methodology for Diffusion Language Models that aligns structural anchors across the entire target response space, replacing prefix prompting. Evaluated on mathematical reasoning, code generation, and trip planning, TI improves performance by 9.40% and accelerates multi-token generation.
PluRule is a multimodal multilingual benchmark for moderating pluralistic communities on social media. It covers 13,371 rule violations across 1,989 Reddit communities (9 languages, 2,885 rules). State-of-the-art vision-language models, including GPT-4.5 with advanced reasoning, only marginally outperform a trivial baseline.
Comparative study of protein representations (amino acid composition, k-mers, ProtBERT, physicochemical descriptors) for Parkinson's disease classification. ProtBERT + MLP achieves F1=0.704±0.028, ROC-AUC=0.748±0.047. Primary sequences alone show limited discriminative power; structural/functional features required.
ReAlignFit, a molecular relational learning method, improves model stability by aligning substructure representations using chemical Induced Fit-based inductive bias. Tested on 9 datasets, it outperforms state-of-the-art models and enhances robustness to distribution shifts (rule-shifted and scaffold-shifted data).
Study on extrinsic gender bias in Bangla pretrained language models. Four manually annotated task-specific datasets constructed (sentiment analysis, toxicity detection, hate speech, sarcasm detection) with minimal-pair gender perturbations. RandSymKL debiasing strategy proposed, combining symmetric KL divergence and cross-entropy loss, reducing bias while maintaining competitive accuracy.
4D video generation model for robot manipulation enforcing multi-view 3D consistency through cross-view pointmap alignment supervision. Generates spatio-temporally aligned video sequences from single RGB-D image per view without camera poses as input. Demonstrates superior visual stability and robot end-effector trajectory recovery on simulated and real-world datasets.
AgentRevive introduces a Markov state-aware framework for resilient multi-agent LLM system evolution. Instead of aggressively pruning failing agents, the method uses soft state transitions (Active/Standby/Terminated) with a hallucination risk estimator. Results: outperforms baselines on general reasoning, domain-specific tasks, and hallucination challenges while reducing token consumption.
arXiv survey on trustworthiness of RAG (Retrieval-Augmented Generation) systems. Proposes Trust-RAG Compass framework assessing 6 dimensions: factuality, robustness, fairness, transparency, accountability, privacy. Introduces TRC Bench evaluation benchmark comparing proprietary and open-source models.
GVGAI-LLM is a video game benchmark for evaluating spatial reasoning and problem-solving in LLMs. Built on the General Video Game AI framework, it contains 118 ASCII games testing planning and logical reasoning. Zero-shot evaluations reveal persistent limitations of current models in spatial reasoning, partially improved by structured prompting.
Novel adversarial imitation learning algorithm combining off-policy learning with double Q-network stabilization. Reduces sample inefficiency of GAIL by eliminating on-policy algorithm dependency (TRPO) and reward engineering requirements.
New framework to stabilize temporal predictions in surgical phase recognition. Introduces TEC loss (training), EGTP (inference), and TFI (metric). Reduces prediction fragmentation on Cholec80 and AutoLaparo while maintaining frame-wise accuracy.
SEED is a framework representing experimental conditions as typed actor-flow graphs to study multi-agent systems and human-AI workflows. It enables describing conditions, evaluating structural novelty, and generating candidate designs under constraints. Empirical test on medical-triage task shows SEED-guided designs provide clearer interaction changes, assumptions, and governance checks.
Researchers apply Multi-Object Tracking (MOT) to camera-trap data to improve wildlife species classification. By fusing softmax probabilities across temporal trajectories, the method gains 5.1% weighted F1-Score on best-performing MOT models, eliminating inconsistent predictions across consecutive frames.
Replay-based continual learning method for adapting pneumonia detection models across clinical domain variations without catastrophic forgetting. Incorporates class-aware balanced replay and dynamically reweighted class-imbalance loss. Achieves 88.66% accuracy on PneumoniaMNIST with 5 simulated domains, outperforming Experience Replay and Fine-Tuning baselines.
ConceptAgent, a training-free multi-agent framework, bypasses concept erasure in diffusion models by exploiting denoising dynamics. The black-box approach awakens suppressed concepts by initializing the denoising trajectory from surrogate-guided noisy states, without access to model parameters.
BM25 improvement for code retrieval using q-logarithmic transformation of RSJ-odds IDF. On CoIR CodeSearchNet Go, NDCG@10 rises from 0.2575 to 0.4874 (+89.3%). Drop-in fix with no latency cost, parameterized by corpus hapax density.
TTE-Flash replaces explicit Chain-of-Thought traces with latent think tokens to accelerate reasoning-aware multimodal representations. TTE-Flash-2B outperforms explicit-CoT counterparts on MMEB-v2 while maintaining constant inference cost. Latent tokens remain interpretable both textually and visually.
New IBAL method to strengthen MARL robustness against inter-agent interaction disruptions. Framework uses information-theoretic approach to construct attacks that degrade coordination by perturbing observations and actions, then trains agents to remain reliable. Demonstrated improvement over existing baselines and agent-missing scenarios.