Model releases, architecture work and prompting research on large language models — from frontier-lab announcements to the arXiv papers behind them.
arXiv:2609.15997v1 Announce Type: cross Abstract: Improving safety at intersections requires identifying crash mechanisms and recommending appropriate countermeasures. However, this process tradition...
arXiv:2609.16689v1 Announce Type: new Abstract: Large-scale vision-language models (VLM) such as CLIP enable strong open-vocabulary reasoning, yet deploying these capabilities on resource-constrained...
arXiv:2609.17269v1 Announce Type: new Abstract: Multimodal large language models generate natural-language responses from visual inputs, yet may mention objects absent from an image. In medication as...
arXiv:2609.17443v1 Announce Type: new Abstract: Vision-language models (VLMs) achieve strong visual question answering (VQA) performance, but processing large cluttered images is computationally expe...
The paper investigates how large language models (LLMs) generate distractor answers for multiple‑choice questions (MCQs) by modeling student misconceptions. It introduces a learning‑science‑based taxonomy of reasoning strategies and applies it to LLM‑generated reasoning traces in math and science MCQs. The study finds that in math, LLMs often follow a misconception‑based process that can be diagnostically useful, whereas in science they rely more on semantic similarity, with frequent failures when the model cannot produce a correct solution or discards plausible distractors. Providing the correct solution in the prompt improves alignment with human distractors by 6.4%. "whyItMatters":"The findings show that anchoring distractor generation to the correct solution enhances LLM alignment with human‑authored distractors, underscoring the importance of correct‑answer cues in educational AI."
The paper "LLM Inference in a Flash!" proposes an integer‑only quantization scheme and a dictionary‑based KV cache compression technique to enable large language model inference on compute‑in‑flash (CIF) devices. By eliminating floating‑point operations and reducing KV cache traffic through sparse dictionary coding, the authors achieve minimal accuracy loss while cutting dynamic KV cache traffic by 15× on Llama‑3.1‑8B and Qwen‑2.5‑7B models.
The paper introduces ViD, a vision‑dominant gender bias mitigation framework for large vision‑language models. ViD uses causal analysis of attention patterns and dual mechanisms—backdoor adjustment and refined token selection—to suppress bias while preserving reasoning and generation quality. Experiments show a 14.7% reduction in gender bias on FACET and significant improvements on MS COCO image captioning, all without extra training overhead.
The paper introduces a new parameter pruning scheme based on the Fisher information metric, using the geodesic distance in model space to determine the minimal change when a parameter is set to zero. This approach creates a hierarchy of pruning methods, starting from traditional magnitude pruning and progressing to more sophisticated, effective schemes. Experiments on fully‑connected networks and vision transformers across MNIST and CIFAR‑10 show that the Fisher‑based method outperforms magnitude pruning and local Fisher pruning in accuracy and Matthews correlation coefficient, while intermediate approximations offer computational efficiency with near‑optimal performance.
The paper introduces Affect-Prototype Guided Fusion (APCF), a framework for open‑vocabulary multimodal emotion recognition that handles incomplete and unsynchronized modal data. APCF builds an affect‑prototype library to model how different emotions contribute across modalities, enabling dynamic fusion of available features. The fused representations are then decoded by an LLM to generate open‑vocabulary emotion labels, achieving superior performance on OV‑MERD+ and MER‑FG datasets compared to existing methods.
FlexEE is an early‑exiting framework designed for large language model inference that is constrained by computation and memory, particularly in offloading‑based deployments. It uses layer‑wise exit supervision, self‑speculative decoding over a Top‑K local vocabulary, and dynamic hidden‑state management to enable reliable intermediate‑layer predictions and memory‑aware execution. Experiments on Llama2‑7B and Llama3‑8B show that FlexEE achieves significant speedups—up to 1.27×/3.16× and 1.25×/2.83× respectively—while maintaining minimal accuracy loss.
The paper introduces Symbolic Separation, a method that grounds deep learning agents in knowledge graphs to improve reliability in operational data analytics. By restricting agent actions to an ontology-constrained Virtual Knowledge Graph with deterministic pre-execution validation, the approach transforms complex queries into validated graph traversals rather than relying on LLM-inferred joins. In experiments on 49.9 TB of supercomputer telemetry, the Neurosymbolic Deep Analyst achieved an 86% task‑success rate, eliminated silent data‑integrity errors, and reduced token costs by 2.4× compared to a non‑symbolic baseline.
The paper presents a method for compressing large tabular foundation models into compact feed‑forward students for use in interactive agentic decision loops. By distilling a TabPFN teacher into a student, the authors reduce model size from 53.2 M to 8,546 parameters (6,220× compression) and from 111.4 M to 17,059 parameters (6,532× compression) while preserving 95.4–100.5% accuracy and 96.8–100.0% AUC across several benchmarks. The study demonstrates that soft‑target distillation yields a 2.1–7.0 AUC improvement over hard‑label training.
The paper examines why learned gates in sparse attention models offer little advantage over random gates when jointly trained with the transformer. Through experiments on a 31M-parameter transformer, the authors attribute this to routing absorption, where the model’s representations adapt to the imposed mask, diminishing the benefit of learned routing. They also explore hard masking, stochastic mask training, and the impact of trainable attention layers on gate performance, concluding that freezing the model stabilizes routing targets for effective post‑hoc sparsification.
arXiv:2609.16340v1 Announce Type: cross Abstract: Machine translation systems are periodically upgraded to stronger models, but the available preference signal is human post-edits of an older system'...
arXiv:2507.23248v2 Announce Type: replace-cross Abstract: Bengali is spoken by more than 230 million people, yet no standardized instrument evaluates large language models (LLMs) on Bengali across th...
arXiv:2609.15998v1 Announce Type: new Abstract: Every large language model (LLM) has behavioral traits and moral preferences that comprise its character. Whether by design or as an emergent property...
arXiv:2609.16010v1 Announce Type: new Abstract: The complexity of legal language and limited accessibility to legal information pose significant challenges to justice delivery in Nepal. Traditional l...
arXiv:2609.16739v1 Announce Type: new Abstract: Background: Large language models (LLMs) have achieved physician-comparable performance on multiple-choice medical knowledge examinations, but their ca...
arXiv:2609.16967v1 Announce Type: new Abstract: Ensuring that multilingual language models generate coherent text in a specific target language is a major issue in multilingual language modeling. We...
arXiv:2609.16006v1 Announce Type: cross Abstract: Large language models (LLMs) increasingly serve users whose expectations are shaped by their cultural context, yet most cultural evaluations test wha...