Model releases, architecture work and prompting research on large language models — from frontier-lab announcements to the arXiv papers behind them.
Autoregressive image generation has emerged as a paradigm for multimodal AI systems due to its compatibility with transformer-based LLM serving infrastructures. However, generating thousands of visual...
MiX: Micro-Inverted-Scaling for End-to-End Low-Bit Vision-Language Model Acceleration proposes a new quantization format that inverts the traditional microscaling approach by assigning private exponents to each element and a shared mantissa. The adaptive dual-format MiX-MX inference framework maps this format to a custom accelerator, replacing multipliers with shifters. Evaluations show that 4.5-bit MiX matches or surpasses NVFP4 accuracy on multimodal benchmarks while improving area efficiency by 25% and delivering 2.3–4.5× speedup with 1.4–2.9× energy reduction compared to the Focus accelerator.
The paper draws a parallel between AI model development and population genetics, treating successive model generations as analogous to sexual and asexual reproduction. It demonstrates that training models on peers’ outputs reproduces classic genetic processes such as the Wright–Fisher model, while combining parents’ weights can either cancel or preserve inherited advantages depending on the method. Experiments across recurrent, feedforward, and language models confirm these analogies and reveal architecture‑specific biases, including the Fisher‑Muller effect and reproductive isolation when lineages learn conflicting conventions.
CompileRover is a new optimization framework for virtual machine compilers that uses a tri‑role LLM‑driven collaboration mechanism involving a referee, an advisor, and an operator. It tackles common issues such as redundant computations, inefficient loops, and suboptimal function implementations by applying control‑flow analysis, code‑structure transformations, and dynamic execution pattern recognition. Benchmark evaluations show that CompileRover consistently outperforms existing virtual machine compilers, reducing execution overhead, improving data‑flow consistency, and enhancing overall compiler performance.
The paper examines the role of schema linking in Text-to-SQL systems and finds that recent large language models can effectively use relevant schema elements even when many irrelevant ones are present. Consequently, the authors eliminate schema linking when the entire schema fits within the model’s context window, instead employing augmentation, selection, and correction techniques to enhance accuracy. Their approach achieves first place on the BIRD benchmark with a 71.83% accuracy.
QiT is a Quantum‑Inspired Transformer designed for visual recognition tasks. It replaces quantum neural network concepts with scalable classical operations: angle‑inspired encoding of image tokens, self‑attention over periodic features approximating quantum fidelity kernels, and gated multiplicative emulation of variational circuit interactions. The model achieves competitive performance on image‑classification benchmarks, matching a classical Transformer while avoiding the high runtime costs of simulated quantum models.
The paper presents a lightweight CNN‑integrated Compact Convolutional Transformer (CCT) designed for multi‑scale feature learning in breast cancer mammography. With only 250,435 parameters, the model achieved 99‑100% accuracy across three datasets using 5‑fold cross‑validation, demonstrating robust generalization. Explainable AI components were added to clarify the classification process, aiming to increase clinical trust in resource‑constrained settings.
The paper introduces ${M}^2$Tok, a Multi-head Multi-codebook Action Tokenizer that reduces reconstruction loss in discrete action tokenization for Vision‑Language‑Action models. By decomposing latent action features into multiple heads and assigning independent codebooks to each, the tokenizer expands representational expressivity and improves policy performance. Experiments on RoboTwin, Simpler‑Env, and zero‑shot real‑world tasks show superior reconstruction fidelity and higher success rates compared to prior methods.
The paper introduces two new tokenisation algorithms—BottomUpLL and TopDownComp—to systematically explore the 2x2 design space defined by optimisation objective (compression vs. log‑likelihood) and search procedure (bottom‑up merging vs. top‑down pruning). Experiments across model sizes, vocabularies, and domains show that the search procedure, rather than the objective, consistently yields lower bits‑per‑byte, while no clear pattern emerges on the BLiMP benchmark. These findings clarify how tokeniser design choices influence language‑model performance and provide guidance for constructing tokenisers more principledly.
Mem2Ego introduces a vision‑language model for embodied navigation that combines global memory with egocentric visual inputs. By adaptively retrieving task‑relevant cues from a global memory module and aligning them with local perception, the framework improves spatial reasoning and decision‑making over long horizons. The method outperforms prior state‑of‑the‑art approaches on the HSSD and HM3D benchmarks and shows strong performance on a real robot.
The tutorial titled "Deep Learning for Sequential Decision Making under Uncertainty: Foundations, Frameworks, and Frontiers" explores how modern deep learning techniques—such as neural networks, transformers, large language models, and deep reinforcement learning—can be integrated with operations research and management science to address complex, uncertain, and dynamic decision problems. It argues that deep learning should complement, not replace, optimization, offering adaptability and scalable approximation while OR/MS provides rigorous constraint and uncertainty modeling. The tutorial organizes the field around predict‑then‑optimize, decision‑aware learning, constraint‑aware decision generation, and deep reinforcement learning, and highlights applications across supply chains, healthcare, energy, and autonomous systems.
The paper audits vision‑language models (CLIP) for gender bias using 1,500 artworks from the Metropolitan Museum of Art, focusing on zero‑shot logit differences for prompts like "masterpiece," "quality," and "influence." Unadjusted results show no significant gender effect and statistical equivalence across models, but high residual variance suggests that global zero‑shot metrics are largely noise‑dominated and may miss fine‑grained biases. The study underscores the need for multivariate confound control, equivalence testing, and provenance auditing when evaluating AI fairness in cultural heritage data.
The paper reports that in English all‑words word sense disambiguation (WSD), the scarcity of high‑quality labels—not the models—has become the limiting factor. The authors introduce lexEN, a human‑adjudicated correction layer over the Maru2022 ALL_NEW benchmark, and SenseBench, a living leaderboard for LLM WSD evaluation. They show that frontier large language models reach about 95 % accuracy on lexEN‑v1, that relabeling corpora with these models improves downstream systems, and that fine‑grained WordNet senses are often ill‑posed, with coarsening improving both annotator agreement and model performance. "whyItMatters":"The study highlights that improving label quality and managing annotation costs are now the critical challenges for advancing WSD performance, as model accuracy is already near its theoretical ceiling."
M‑SQE is a post‑retrieval framework that estimates the quality of multilingual agent skills by combining a Theory view (intrinsic quality) and an Action view (task‑grounded utility) into a domain‑conditioned score. It was evaluated on general, tool‑use, and cultural skill‑use domains, showing a task‑success improvement of at least +3.5 points over baselines across three retrievers. The method notably boosts performance for low‑resource languages, raising Hindi by +12.9 pp and Swahili by +5.6 pp, and achieves strong results across six cultural regions, advancing linguistic and cultural equality in agentic skill use.
PersonaPath is a new benchmark for knowledge‑centric personalized learning path planning, pairing 2,000 learner personas with a hierarchical knowledge graph of 347 textbooks, 1,751 units, and 4,092 concepts across 77 subjects. The study evaluates large language models on this benchmark, finding that even the best model achieves only a 29.5% final pass rate in Basic Education and fails to exceed 44.7% in tailoring paths to individual learners, highlighting a significant adaptivity gap. This work underscores the challenge of moving beyond exercise‑centric recommendation toward goal‑oriented, curriculum‑scale guidance.
The paper introduces PanicCognitivePath (PCP), a model that predicts the timing of panic emotional arousal by integrating appraisal emotion theory into a Belief‑Desire‑Emotion‑Intention (BDEI) framework. PCP uses a Psychological Safety Distance (PSD) model to fuse physical, social, cognitive, and informational signals into a unified risk metric, and confines large language models to a single parameter‑estimation step to reduce hallucinations. Experiments on Hurricane Sandy data show PCP improves individual prediction accuracy by 10.68% and reduces peak count error to 7.07%.
The paper introduces ASLEval, a framework for measuring privacy exposure displacement in large language model (LLM) agent sessions. It highlights that traditional local proxies—such as inspecting a single action or final response—often miss unauthorized data leaks elsewhere in a multi-step session. ASLEval pre-registers hidden target sets, tracks all declared visible exits, and preserves internal traces for diagnosis, revealing that a single outlet view can overlook nearly 47% of exposure and that internal evidence typically precedes visible leaks. The study underscores the need for benchmarks that define complete visible boundaries, ground claims in pre-specified targets, and report privacy alongside task utility.
The study evaluates whether vision‑language models (VLMs) can reliably assess sidewalk accessibility attributes—effective width, longitudinal slope, cross slope, and pavement condition—from pedestrian‑level images. Using sampling‑based conformal prediction on 514 images from Seoul, the authors find that calibrated models achieve nominal 90% coverage, but only effective width yields informative estimates; other attributes remain too uncertain for compliance assessment. The work also demonstrates that raw sampling dispersion is not a trustworthy uncertainty measure without calibration and releases annotated images with ground‑truth measurements.
Agora is a system that uses Git as a shared memory for autonomous research agents, recording each claim as an immutable commit in an append‑only directed acyclic graph. In a 12‑day run, 13 language‑model workers independently explored a weight‑transfer problem, producing 1,703 contributions that improved a 119.6M‑parameter model’s performance from 3.39 to 1.899 bits per byte. The system’s design includes a diversity‑aware selection rule and an index that tracks the frontier, neglected branches, and verification status of each claim.
LIGE‑GR is a framework that transitions traditional ranking‑based recommender systems to a generative, listwise approach inspired by large language models. It extends existing pointwise recommendation models into a listwise generation system, enabling sequence‑level optimization without overhauling the entire infrastructure. Experiments on Instagram Reels and Facebook Video show modest gains in user time spent—1.14 % and 0.72 % respectively—while adding only slight inference overhead.