Leaderboards, eval harnesses and ablations — the contested business of deciding which model is actually better.
arXiv:2510.17904v3 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are widely used because they process structures, syntax and code well, but this same ability also makes them par...
arXiv:2512.07624v2 Announce Type: replace Abstract: Process Model Forecasting (PMF) aims to predict how the control-flow structure of a process evolves over time by modeling the temporal dynamics of...
arXiv:2601.01352v2 Announce Type: replace Abstract: Human identity-preserving text-to-video generation remains challenging under large changes in viewpoint, facial expression, illumination, and motio...
CCTU is a new benchmark designed to evaluate large language models (LLMs) on their ability to use tools under complex constraints. It includes 200 test cases that average seven constraint types and 4,700‑token prompts, covering resource, behavior, toolset, and response dimensions. An executable validation module performs step‑level checks, and nine state‑of‑the‑art LLMs were tested, revealing that none exceed a 20% task completion rate when strict constraints are enforced, with frequent violations and limited self‑refinement.
arXiv:2604.04720v2 Announce Type: replace Abstract: Large Reasoning Models (LRMs) still exhibit large performance gaps between English and other languages, yet much current work assumes these gaps ca...
arXiv:2604.24952v2 Announce Type: replace Abstract: Human visual preferences are inherently multi-dimensional, encompassing aesthetics, detail fidelity, and semantic alignment. However, existing data...
arXiv:2605.01699v4 Announce Type: replace Abstract: Recent attacks show that behavioural unlearning of large language models leaves internal traces recoverable by adversarial probes. We characterise...
arXiv:2605.11217v2 Announce Type: replace Abstract: The model context protocol (MCP) has been widely adopted as an open standard enabling the seamless integration of generative AI agents. However, wh...
SiST‑GNN introduces a simultaneous spatial‑temporal message‑passing framework for dynamic graph neural networks, fusing per‑node temporal embeddings with spatial aggregation in a single operation. By maintaining a recurrent hidden state per node and treating it as a cross‑time edge, the model jointly reasons over topology and evolution. Experiments on link‑prediction and node‑classification benchmarks show significant improvements over prior methods, achieving up to 158% gains in live‑update link prediction and outperforming discrete‑time baselines by 7–23% in dynamic node classification.
arXiv:2608.16353v2 Announce Type: replace Abstract: Internal-state probes enable truthfulness prediction before a large language model generates an answer. When detectors change both the layers they...
PAGE is a partition‑aware gated KV‑cache eviction method that reframes eviction as a per‑input admission decision. It uses a single label‑free scalar— the early‑to‑late drop in pairwise top‑k head agreement—to classify inputs into a capacity‑bound class (where eviction is catastrophic) and a dilution‑prone class (where eviction is safe or beneficial). By thresholding this drop, PAGE applies a base evictor only when necessary, reducing the harm rate in the capacity‑bound regime from 0.75 to 0.026 and achieving a 29× improvement across four models and benchmarks without retraining the evictor.
The paper introduces a hybrid attention model that learns a unified time‑aware patch representation for irregular multivariate time series (IMTS) forecasting. It employs a time‑aware patch encoding to embed variable‑length intra‑patch timestamps, a time bias attention mechanism to adjust for temporal misalignment and asynchronous cross‑channel dependencies, and a hybrid causal mask on a decoder‑only Transformer to balance historical context with autoregressive forecasting. The authors also curate VersaTSA, a 30 B‑observation dataset preserving native sampling sparsity, and demonstrate state‑of‑the‑art zero‑shot performance on three IMTS benchmarks while remaining competitive on regular MTS tasks.
arXiv:2607.22068v2 Announce Type: replace-cross Abstract: Multi-branch architectures and CNN-Transformer fusion are widely believed to improve vehicle re-identification (Re-ID) by combining complemen...
arXiv:2609.23184v1 Announce Type: new Abstract: Embodied world models learn to predict future physical dynamics from visual observations and control signals, where physical knowledge is implicitly en...
arXiv:2609.24337v1 Announce Type: new Abstract: While Mamba-based models have shown strong potential for long sequence modeling, adapting them to vision is challenging due to the requirement of local...
The paper introduces Regularized Recursive Self-Improvement of Agent Harnesses (RRSI), a method that applies regularization principles to the iterative editing of an LLM agent’s harness—prompts, control flow, tooling, memory, and context management. RRSI limits the number of edits per candidate, encourages novel trajectories, and uses a critic and pruner to filter out benchmark‑specific or ineffective changes, thereby favoring reusable agent mechanisms. Experiments on eight benchmarks show RRSI improves performance by up to 14.1 points on the training split and 4.7 points on out‑of‑distribution tests, while reducing policy token usage by 30% compared to unregularized evolution.
The paper introduces MedGym, a benchmark environment for dynamic medical treatment recommendation that models patient evolution in continuous time using Physics-Informed Neural Networks. It addresses gaps in existing reinforcement learning (RL) approaches by allowing evaluation of RL methods under irregular measurement intervals, personalized treatment responses, and safety considerations. MedGym enables direct comparison between discrete-time and continuous-time RL methods and supports clinically relevant metrics such as personalization and trajectory-level safety.
The study evaluates large language models (LLMs) on sequential emergency department triage, where acuity labels are predicted from progressively longer nurse‑patient conversations. Six LLMs were tested at five checkpoints on simulated and physician‑authored dialogues, showing a decline from moderate‑to‑substantial agreement on full records to only fair‑to‑moderate agreement at each checkpoint. The models consistently anchor on chief complaint exchanges and fail to integrate later evidence, yielding low agreement with clinicians (QWK 0.295 vs. 0.887‑0.929) and concentrating predictions on ESI‑2 and ESI‑3. whyItMatters":"The findings reveal that LLMs, despite strong offline performance, cannot reliably handle the sequential nature of real‑time triage, highlighting a critical gap for safe deployment in emergency settings."
The paper introduces a Channel‑Boosted Multi‑Agent System (CB‑MAS), instantiated as IC‑MAS, to classify document sensitivity without the input‑length truncation problem of standard transformers. IC‑MAS uses a Channel Critic Agent to adaptively weight two first‑window encoders and Consultation Agents to exchange belief states, achieving 90.72% accuracy and 91.23% F1‑score while reducing computation by ~54% compared to a fixed‑round baseline. The authors provide explainability via LIME/SHAP, multi‑agent evaluation, and a transparent discussion of limitations.
arXiv:2603.18532v3 Announce Type: replace-cross Abstract: The strong performance of large vision-language models (VLMs) trained with reinforcement learning (RL) has motivated similar approaches for f...