Model releases, architecture work and prompting research on large language models — from frontier-lab announcements to the arXiv papers behind them.
AiSearch is a flexible multimodal retrieval framework that uses Vision Language Models (VLMs) to enable natural language search over images and videos. It supports interactive search refinement through user feedback, allowing results to be tailored to the user's intent in real time. The system also provides visual benchmarking across multiple VLMs, enabling users to choose the most suitable model for their specific task.
Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States introduces PoS, an inference-time framework that builds and maintains explicit belief states to guide large language model agents. Each belief state combines an estimate of the current world with unresolved task requirements, making clear what the agent still needs to learn and accomplish. PoS validates consistency, monitors task progress to detect Belief Trapping, and tailors recovery to the trapping pattern and unresolved requirements, achieving top performance across four benchmarks with all three LLM backbones.
The paper introduces MCRI, a four‑dimensional framework for analyzing and evaluating agent skills, and implements it as MCRI‑Eval, a large‑language‑model‑based evaluation method. Using 63,812 public skills from OpenClaw and 58,275 skill‑conditioned executions across BigCodeBench, BFCL‑Fundamental, and Mind2Web, MCRI‑Eval’s scores correlate with community popularity and outperform other methods in downstream ranking agreement. The evaluation also shows that MCRI‑Eval improves top‑1 skill selection by 17.7, 22.8, and 19.6 percentile points on the three benchmarks, offering a valuable pre‑execution signal for prioritizing promising skills before costly evaluation.
The paper argues that evaluating agents as fixed models is insufficient, proposing instead to treat them as configurable systems. Using a new benchmark of four scientific tasks, the authors analyze how five configuration aspects—task information, reasoning, self‑verification, time budget, and backbone model—affect performance, noting that about 54% of outcome variance arises from run‑to‑run differences even with the same settings. The study finds that providing more task information has the strongest impact, while interactions among settings (e.g., extra time only helps with adequate information or model capability) and the choice of verification tools significantly shape agent behavior.
ScholarCatalyst is a new benchmark that evaluates how well AI systems can retrieve research papers that inspire new work. The dataset was created by having 184 lead authors of 207 recent computer science papers annotate which earlier papers helped their projects, providing detailed rationales. The benchmark tests retrieval from the literature available at the start of a project, revealing that current agentic search and even advanced models like Claude Fable 5.1 perform only modestly better than simple embedding retrieval.
The paper introduces a bounded‑fidelity sim‑as‑demo‑stage design pattern that suppresses contact physics during object handoffs in simulators, using MuJoCo’s mocap‑body primitive and a lightweight Python adapter. This approach ensures audit‑chain stability, producing identical event‑log hashes across 1,000 replays per posture, whereas a contact‑force baseline yields significant divergence. The authors demonstrate that the pattern maintains reproducibility across various timesteps and sequential handoffs with minimal overhead, and they identify contexts where it should not be applied.
ShatterQuant is a hardware-software co-designed framework that enables mixed-precision quantization within individual tensors by assigning different bit-widths to blocks of a weight projection. It couples precision granularity with processing element configuration, allowing each precision to determine an effective block height. The framework includes a hardware-aware post-training method based on block-level standard deviation and weight sensitivity, a ShatterQuant Transformer Accelerator supporting 1/2/4/8-bit weight precision, precision-dependent PE configuration, block rescaling, and integrated softmax and piecewise-linear nonlinearities, and an evaluation showing 1.5 TOPS, 760 GOPS/$mm^2$ area efficiency, and 2.8 TOPS/W energy efficiency on a TSMC 16nm PDK implementation.
LENS‑GRF is a permutation‑invariant lesion evidence network that uses a Set‑Transformer and gated residual fusion to combine global facial context with localized lesion patches for four‑class acne severity grading. The framework integrates adaptive facial skin segmentation, a Vision Transformer prior, and a lesion set transformer that encodes spatial geometry, with a gating mechanism that modulates local residual contributions. In experiments on ACNE04 and PLSBRACNE01, the fully automated model achieved 80.82% accuracy, while using ground‑truth lesion annotations raised accuracy to 95.89% and a Quadratic Weighted Kappa of 0.9753; zero‑shot evaluation on the full cohort yielded 35.00% accuracy versus 42.50% for a global baseline, and oracle analyses on a 148‑subject cohort showed improved accuracy and QWK up to 47.97% and 0.5799.
The paper introduces a new approach to distill reasoning abilities from large language models (LLMs) into smaller student models by framing the task as a constrained reinforcement learning problem. It enforces a worst‑case constraint on the teacher’s log‑likelihood for every prefix of the reasoning chain, avoiding reward hacking and excessive teacher regularization. Experiments on mathematical reasoning and code generation show that this method improves the balance between accuracy and fidelity, achieving the highest rigorous reasoning success rate among evaluated settings.
STCFormer is an adaptive spatio‑temporal Transformer that dynamically clusters weather stations within each temporal patch, combining fine‑grained local attention inside clusters with global attention over regional summaries. The model’s design is supported by a Lipschitz upper bound that suggests robustness benefits, and it achieves the lowest 24‑hour mean squared error across eight temperature and wind forecasting tasks on three real‑world datasets, ranking first or second in 47 of 48 comparisons. Ablation studies and case analyses confirm the advantages of locally adaptive grouping and complementary local‑global interactions.
arXiv:2610.00385v1 Announce Type: cross Abstract: Replay selectors often rank cached trajectories by format feedback, confidence, freshness, or response length, although cache-level correctness and d...
arXiv:2610.00388v1 Announce Type: cross Abstract: Reinforcement learning enables large language model (LLM) agents to learn multi-step behaviors through interaction with their environments. However,...
arXiv:2610.00400v1 Announce Type: cross Abstract: Multi-turn attacks on agentic systems can compose individually permissible actions into harmful outcomes, challenging defenses that assess actions or...
IrekoGPT is a post‑hoc technique that transforms pretrained large language models into slimmable versions, enabling dynamic width adjustment during inference. It builds on SliceGPT by keeping the original projection matrices intact, thereby exposing nested subnetworks at various widths. The method enhances robustness through layer‑wise calibration across multiple compression ratios and refines downstream linear layers using gradient‑free ridge regression, yielding better performance than naive PCA‑based slimming on Llama and Qwen models, especially at high compression levels.
arXiv:2610.00432v1 Announce Type: cross Abstract: Trellis-coded quantization enables high-dimensional compression of large language model (LLM) weights at ultra-low bit widths without the exponential...
arXiv:2610.00497v1 Announce Type: cross Abstract: We present Gumbel Straight Flow (GSF), a continuous flow map language model that leverages the noise-data coupling of a pretrained autoregressive lan...
arXiv:2610.00620v1 Announce Type: cross Abstract: Grokking refers to the delayed emergence of validation-set generalization after a model has already overfit the training set. Although first observed...
arXiv:2610.00675v1 Announce Type: cross Abstract: Evolutionary approaches to LLM-driven discovery often generate new programs from a small set of selected ancestors. This keeps contexts manageable bu...
arXiv:2610.00814v1 Announce Type: cross Abstract: Synthetic data are increasingly used to scale LLM training, yet more synthetic data do not necessarily produce better models. Useful synthetic data m...
arXiv:2610.00838v1 Announce Type: cross Abstract: Agentic reinforcement learning (RL) trains a large language model (LLM) to act over long, multi-step interactions. However, a single localized error...