Machine Learning News

Researchers have developed IN2R, a new framework that corrects mismatched image-text pairs in large web-harvested datasets by synthesizing continuous supervision signals from intra-modal data relationships rather than re…

Researchers introduced Learnable Rank LoRA (LR-LoRA), a parameter-efficient fine-tuning method that allows the adapter rank to be learned during training rather than fixed, enabling the optimizer to determine the appropr…

Researchers have developed a consequence-aware test-time compute allocation method for AI reasoning models that prioritizes tasks based on the real-world cost of errors rather than predicted difficulty alone. The approac…

A new study from arXiv revisits eight memory systems for large language model agents, finding that most fail to generalize across diverse deployment scenarios like single-turn QA and long-horizon tasks. The researchers i…

Researchers have developed Deliberate Evolution (DE), an agentic framework that improves sample efficiency in symbolic regression by decoupling candidate proposal from search guidance. The method uses adaptive operators,…

Researchers introduced Curation-Bench, a benchmark testing whether generalist coding agents can automate the labor-intensive process of curating AI training data. Out-of-the-box agents matched strong published data-selec…

Researchers have developed StepPRM-RTL, a framework that enhances LLM-based RTL code generation for digital hardware design by combining stepwise trajectory modeling, process-reward modeling, and retrieval-augmented fine…

Researchers have introduced VAMPS (Visual-Assisted Mathematical Problem Solving), a benchmark of 1,168 bilingual multiple-choice questions designed to test whether multimodal AI models can improve their reasoning by cons…

The IEEE P3109 draft standard defines a new family of parameterized binary floating-point formats designed for machine learning, allowing efficient value representation in few bits with adjustable width, precision, signe…

A new pseudospectral theory for block-triangular Jacobians in coupled gradient descent proves that transient amplification before convergence can be arbitrarily large even when asymptotic stability is guaranteed, with th…

A systematic study of query, key, and value (QKV) projection variants in Transformers found that sharing the key and value projections (Q-K=V) performs on par with or better than the standard three-projection approach ac…

Researchers introduced SDPG, a self-distilled policy-gradient framework that combines group-relative verifier advantages with normalized standard deviation and full-vocabulary on-policy self-distillation to improve reinf…

Researchers have developed LiftQuant, a framework enabling continuous bit-width control for large language models by using a "lift-then-project" mechanism that approximates low-dimensional weight vectors from a higher-di…

Researchers developed an XGBoost machine learning model using eight routine clinical features from the Alzheimer's Disease Neuroimaging Initiative dataset that achieved 94.3% accuracy and a 0.982 macro AUC in classifying…

Researchers introduced Stein kernelized molecular dynamics (SKMD), an enhanced sampling method that uses interacting particle dynamics to acquire informative training configurations for active learning of machine learnin…

A new position paper argues that reinforcement learning systems deployed in real-world applications should operate as continual learning agents rather than following the current train-then-fix paradigm. The authors ident…

Researchers have introduced DLLG (Dynamic Logit-Level Gating), a framework that learns token-level fusion weights for combining multiple specialized large language models without requiring token-level labels or expert re…

A new arXiv paper presents a data-centric taxonomy of 3D vision, connecting geometric representations, datasets, learning frameworks, and applications into a single conceptual map. The work analyzes principal 3D data str…

Researchers have developed a methodology for the inverse design of critical experiments using deep neural network surrogate modeling and gradient optimization to maximize neutronic similarity to target nuclear technologi…

Researchers have developed UniCanvas, a diffusion-based unified model that generates interleaved text and images within a single pixel canvas. The model embeds language as visual patterns during image synthesis, overcomi…