Model releases, architecture work and prompting research on large language models — from frontier-lab announcements to the arXiv papers behind them.
arXiv:2609.30988v1 Announce Type: new Abstract: Real-world image super-resolution (SR) requires recovering perceptually realistic high-resolution images from complex low-resolution observations while...
arXiv:2609.31451v1 Announce Type: cross Abstract: The growing popularity of short videos has driven demand for one-click content creation. Visual templates turn uploaded images into personalized cont...
arXiv:2506.02015v4 Announce Type: replace Abstract: Recent advances in Multimodal Large Language Models (MLLMs) have enabled unified multimodal understanding and generation. However, they still strug...
arXiv:2512.21004v2 Announce Type: replace Abstract: Recent advances in pretraining general foundation models have significantly improved performance across diverse downstream tasks. While autoregress...
arXiv:2604.10766v5 Announce Type: replace Abstract: Open-set 3D macromolecule detection in cryogenic electron tomography eliminates the need for target-specific model retraining. However, strict VRAM...
arXiv:2604.17422v2 Announce Type: replace Abstract: Long video understanding remains a formidable challenge for Multimodal Large Language Models (MLLMs) due to the prohibitive cost of processing dens...
The paper introduces an architectural mediation approach that uses the Model Context Protocol (MCP) to bridge large language model (LLM) agents with data spaces. By implementing the Eunomia Agent, the mediation layer translates data space capabilities into structured, schema-driven tools that LLM agents can discover and invoke while respecting governance constraints. A prototype demonstrates end‑to‑end interaction across catalog discovery, metadata retrieval, and data service invocation without altering existing data space components, showing that protocol‑based mediation enables interoperable, standards‑aligned integration of AI agents into governed data‑sharing ecosystems.
The paper introduces skill cascading attacks, where a malicious goal is spread across multiple seemingly benign skills, causing harmful outcomes when combined. It presents SkillCascade, an automated red‑teaming framework, and releases SkillCascade‑Bench, a benchmark of 213 validated cascading test cases across various agent systems and domains. Experiments show that these cascaded interactions reliably induce harmful behaviors while evading existing per‑skill scanners and runtime monitors, revealing a gap between component‑level integrity and system‑level safety.
The paper introduces a knowledge‑graph‑based evaluation framework, S3KG, to assess whether large language models truly understand context in question answering tasks. S3KG combines structural and semantic signals into a single similarity score and is paired with a diagnostic analysis that pinpoints reasoning errors at the triplet level. Across nine benchmarks, the method outperforms existing baselines, achieving up to +7.6 F1 points and an AUROC of 0.973.
BioEVAL is a global, multi‑institutional benchmark that evaluates large language and multimodal models on experimental reasoning tasks in bioengineering. It comprises 608 items across 11 bioengineering subfields, including 359 vetted multiple‑choice questions, 218 literature synthesis tasks, and 10 multimodal image‑interpretation problems. The benchmark was used to assess models such as ChatGPT, Gemini, and Grok, revealing accuracies up to 90% on MCQs, a 0.72 similarity score on literature synthesis, and 80% accuracy on multimodal tasks, with notable variation across subfields.
The study evaluates how well language‑model agents can simulate individual social media reactions by comparing predictions under different prompt conditions. Eight Serbian participants’ reactions to 68 posts were recorded, and four language models were asked to predict these reactions using prompts that varied in profile content and instruction style. The results show that prompts emphasizing attitudinal content and intuitive, immediate responses yield the highest fidelity, outperforming demographic backstories and a crowd baseline, and suggesting that such agents could act as general‑purpose simulated users.
Atelier is a self‑supervised framework that uses a transformer‑based hypernetwork to generate implicit neural representations (INRs) for cryo‑EM maps, enabling efficient, scale‑agnostic, coordinate‑conditioned feature extraction. Trained on 5,439 maps from the Electron Microscopy Data Bank, the pretrained INR provides continuous local feature fields that can be used as auxiliary channels for a 3D nested U‑Net, improving voxel‑level property prediction across eight tasks compared to a volume‑only baseline. The approach demonstrates that amortized INRs can serve as a geometry‑aware primitive for large‑scale cryo‑EM analysis.
HARDEN is a constrained evolutionary search method that transforms existing evaluation cases into more challenging variants while preserving their expected outputs. It operates along domain‑specific complexity axes and enforces feasibility constraints such as task semantics, realism, and execution validity. Experiments on FinQA, PubMedQA, and ContractNLI with Qwen3.5 models show that HARDEN can reduce task‑model accuracy by an average of 22.7% and up to 49.9% compared to single‑pass baselines.
T‑RoPE introduces a time‑aware Rotary Position Embedding for sequential recommendation, replacing index‑only rotations with timestamp‑based angles, learnable temporal coefficients, multiscale frequency banks, shifted query alignment, and non‑stationary key rotation. The authors prove that standard RoPE is time‑translation invariant and cannot capture seasonal contexts, while T‑RoPE breaks this invariance while preserving the RoPE interface. Across five public benchmarks and an industrial‑scale e‑commerce dataset, T‑RoPE outperforms baselines by up to 130 % in HR@10 and delivers significant online lift in conversion and order count.
The study investigates whether adding inference-time reasoning to large language models (LLMs) improves trading performance. Using a controlled experiment across DeepSeek, GPT, and Gemini models, the authors varied reasoning effort while keeping other variables constant and evaluated over a full year of U.S. equities under three input conditions. Results show that additional reasoning does not reliably increase net portfolio returns and can even lead to nonmonotonic performance and unstable outcomes.
LAVOIR is a single‑pass decision encoder that not only predicts answers to typed questions but also identifies which missing pieces of information (slots) would most improve its confidence. By placing candidate slots next to answer options, one forward pass yields both the decision distribution and the expected value of asking each slot, without requiring human labels. In controlled experiments, LAVOIR’s question policy matches a greedy oracle and improves accuracy by up to 14.1 points over never asking, while on real conversations it raises accuracy by 8.3 points with minimal questioning.
The paper introduces Learning What to Skip (LW2S), a method that learns when to omit components in multi‑agent LLM workflows by treating omission as counterfactual credit assignment. LW2S builds action‑specific safety models from controlled skip interventions and uses calibration plus domain‑native guards to decide which steps to skip. Experiments on mathematical reasoning, multiple‑choice QA, and code generation show that LW2S cuts token cost while maintaining or improving overall accuracy, and further studies reveal component redundancy and limitations of agreement‑based skip selection.
ORCA is a new benchmark for evaluating large language models on Data Science Code Translation (DSCT), comprising two settings: ORCA-MAIN with 1,600 grounding-level tasks across data querying, manipulation, and deep learning, and ORCA-PROJECT with 200 full-project translation tasks across seven data‑science task types. Each task includes reference translations and test cases to verify functional equivalence, and a multi‑stage quality verification process ensures task correctness. Experiments show that even state‑of‑the‑art LLMs perform poorly on DSCT, with Claude‑Opus‑4.6 achieving only 56.92% success on ORCA‑MAIN and 33.67% on ORCA‑PROJECT, while an intent‑augmented approach improves success rates by 4.80% and 5.33% respectively.
Backbone-Adaptive Evidence Routing (BAER) is a method that dynamically selects an evidence gathering protocol—evidence stacking, reliability-based expert routing, or candidate-blind reference verification—based on the judge backbone and benchmark during development, then locks that choice for testing. BAER maintains candidate symmetry, ensuring that swapping responses reverses preference but not its strength. In experiments across four benchmarks and two 8B judge backbones, BAER outperforms all compared methods, achieving the highest test accuracy in all eight conditions and improving over the strongest baseline by 0.87–7.32 points.
The paper introduces Hyperbolic Clinical Ontology Embeddings (HCOE), a method that transforms frozen BioBERT embeddings into a Poincaré ball to capture medical code hierarchies. HCOE employs ontology-guided contrastive learning and coarse‑to‑fine ontology‑path aggregation, leveraging ICD, CCS, and ATC hierarchies. Experiments on MIMIC‑IV demonstrate superior performance in clinical relation prediction, hierarchy transfer, and various predictive tasks such as mortality, readmission, medication recommendation, and rare drug prediction.