Machine Learning News

A new arXiv preprint (arXiv:2608.13087v1) reports that apparent gains from non-uniform test-time budget allocation in neural combinatorial optimization (NCO) are largely an artifact of in-sample evaluation. Across three …

A study evaluating three open-source vision-language models (VLMs) — InternVL, Qwen-VL, and SmolVLM — on classifying proxemic danger from egocentric robot images found that without fine-tuning, all models perform near a …

Researchers propose SDAM, a structure-difference-aware memory evolution method for complex Text-to-SQL, integrated into the SDAM-SQL framework. SDAM-SQL achieves 2.0 and 0.4 improvement on BIRD-dev and Spider-test compar…

Researchers propose RA-DPO (Reliability-Aware Direct Preference Optimization), which integrates annotator agreement, model confidence, and token-level uncertainty to improve sexism detection. On 6,920 multilingual posts …

Researchers introduced AIM, a framework interpreting momentum as a residual-driven multiplier correction, and proposed RADAR, an optimizer combining relativistic adaptive geometry, decoupled residual correction, and seco…

Researchers introduced Balanced Adaptive Prototype Selection (BAPS), a framework that compresses contexts for TabPFN inference on large-scale tabular data without retraining. Experiments on the million-row HIGGS and SUSY…

Researchers propose LoRA-Diffusion, a parameter-efficient fine-tuning method that applies low-rank decomposition to the denoising trajectory of diffusion-based language models, achieving the highest mean performance on S…

Researchers propose PAC-Bayes-regularized Meta-LoRA, a method for cross-domain LLM personalization that calibrates adaptation to evidence quality, reducing cross-domain win-rate degradation by 47.9% and improving win rat…

Researchers at an undisclosed institution introduced a proxemics-based reward formulation for deep reinforcement learning (DRL) social navigation that models each human's personal space as a radial Gaussian-mixture field…

Researchers propose Structure-Preserving Epistemic Neural Networks (S-PENNs), a framework for uncertainty quantification in scientific machine learning models with hard architectural constraints, and instantiate it for G…

Medical AI is moving beyond static datasets by training on simulated patients, which forces models to navigate uncertainty and iterative reasoning rather than pattern matching. The workflow involves creating digital pers…

MarkTechPost published a tutorial on building a compact reasoning-focused language model by streaming the SupraLabs reasoning corpus from Hugging Face, applying quality filters, and fine-tuning SmolLM2-135M-Instruct with…

RAGFlow, an open-source RAG engine launched in April 2024 by InfiniFlow, has reached 88,000 GitHub stars and continues trending in August 2026, despite repeated declarations that RAG is dead. The project's success stems …

A developer benchmarked Gemini, Claude, and OpenAI models for structured document extraction using LangSmith and LangGraph, finding that Gemini and Claude tied on field accuracy but Gemini achieved more perfect records. …

JD Logistics, the supply chain arm of JD.com, has evolved from an e-commerce delivery network into a technology-driven platform integrating AI, robotics, and automated warehouses. The company uses AI for demand forecasti…

PyTorch's torch._assert_async API enables asynchronous assertion on GPU device streams, avoiding graph breaks in torch.compile. The API accepts a boolean tensor and reports failures only when the GPU stream synchronizes …

X has open sourced a larger version of its For You feed ranking algorithm on GitHub under the Apache License 2.0, including configuration parameters, weights, visibility-filtering code, and label systems, and is piloting…

Applied Intuition, Inc., a $15 billion Silicon Valley company founded in 2017, is hiring a Data Engineer for its Axion product in Sunnyvale, California, with a salary range of $200k–275k/yr. The role involves building op…

A developer has detailed a method for fine-tuning OpenAI's Whisper-v3 model to detect sleep apnea and hypopnea events from audio recordings. The approach repurposes Whisper's speech-to-text architecture to classify acous…

Reinforcement learning (RL) for large language models (LLMs) achieves rapid performance gains despite being information-theoretically inefficient compared to pretraining, according to a speculative analysis. The author n…