Topic
machine learning
Technology Jailbreaking Frontier AI Models Is Cheap and Easy, New Report Warns Enterprise Users
A new report from AI safety nonprofit FAR.AI shows that jailbreaking some of the most advanced AI models is frighteningly easy and cheap—as low as $58 for Grok. The findings highlight the need for enterprise buyers to scrutinize model safety before deployment.
Technology Beijing Accuses US AI Firms of Using Chinese Models for Training
The Chinese commerce ministry accused US artificial intelligence firms of using Chinese models to train their own AI systems through a process called distillation. This comes after US Treasury Secretary Scott Bessent threatened sanctions against China over alleged technology theft. China defended distillation as a widely used industry practice and vowed to take all necessary measures to safeguard its interests.
Supply Chain 70% Logistics Cost Cut? Custom AI Models Promise 70-80% Savings for Supply Chains
FreightWaves reports that Pallet CEO Sushanth Raman advocates for custom AI models that reduce logistics execution costs by 70-80%. The so-called sovereign AI approach also boosts data privacy and ROI, positioning bespoke intelligence as a critical asset for supply chain success.
Technology project44 CEO: AI Agents Without Context Are Just Guessing Faster
project44 CEO Jett McCandless argues that AI agents require rich contextual data to be effective. The company's Agentic Workflow Manager layers first- and third-party agents on top of shipment-level data to automate tasks like LTL dispatch reconciliation, processing 75,000 dispatches daily and matching over 2,000 that would otherwise require manual intervention.
Scientists Use AI and Quantum Computing to Generate New Peptides in Spare Time
Researchers at the Technical University of Denmark used a hybrid AI-quantum computing system to generate novel peptides, achieving better results than classical models especially with limited data. The work, done on weekends with leftover funds, could accelerate personalized immunotherapies and vaccines.
Technology Self-Improving AI Isn't Just for Frontier Labs: How Enterprises Can Build Their Own
A journalist demonstrates building a self-improving AI using tools from Andrej Karpathy's AutoResearch and startup Prime Intellect. The experiment shows that recursive self-improvement is accessible beyond big labs, with implications for enterprises seeking specialized models.
New Graph Neural Network Learns Protein Representations with Secondary Structure and Energy-Filtered Hydrogen Bonds
Researchers propose a secondary-structure-aware graph neural network for protein representation learning. The model augments residue-level node representations with secondary structure assignments and constructs edges from hydrogen-bond interactions filtered by energetic strength. It achieves consistent improvements over existing methods on standard protein benchmarks and offers enhanced biological interpretability.
SoftSkill: Compressing AI Agent Skills into Compact Latent Controls Boosts Accuracy Over Traditional Prompting
Researchers propose SoftSkill, a method that compresses natural-language agent skills into compact continuous vectors, improving accuracy on benchmarks like LiveMath by 42.1 points over no-skill prompting. The approach uses a frozen backbone and a trainable soft delta, offering a more efficient alternative to traditional Markdown skill files.
FineREX Boosts Knowledge Graph Quality by 31% in Human Smuggling Document Analysis
Researchers introduce FineREX, a fine-tuned NER-RE pipeline for knowledge graph construction from unstructured legal documents. Compared to a larger general-purpose LLM baseline, FineREX achieves absolute improvements of 15.50% in entity F1 and 31.46% in relation F1, reduces legal noise by nearly half, and cuts processing time by 50%.
New Research Shows Pretraining Data Composition Can Engineer Neural Scaling Laws for Particle Physics
A new arXiv paper demonstrates that neural scaling laws in particle physics can be engineered by adjusting pretraining data composition. The study shows that including more diverse and task-aligned synthetic data can shift scaling behavior to require more data rather than larger models, offering insights for efficient AI training.
New Machine Learning Framework Creates Global Ease of Living Index for Major Economies
A new study presents a machine learning framework to create a Global Ease of Living Index, combining various socio-economic and infrastructural factors into a composite score for major economies since 1970. The index uses dimensionality reduction to address missing data and aims to help policymakers identify areas for improvement.
Bi-Anchor Interpolation Solver Cuts Generative Modeling Steps from 100 to 10, Researchers Show
Researchers introduce the Bi-Anchor Interpolation Solver (BA-solver) for accelerating flow matching generative models. It achieves quality comparable to 100+ step solvers in just 10 steps, using a small SideNet (1-2% of backbone size) and novel bidirectional temporal perception. The method is plug-and-play with existing pipelines.
Hard or Just Unreached? Diagnosing the Sampling Blind Spot in Math-Reasoning Difficulty Estimation
A new paper shows that pass@k, the standard metric for estimating math-reasoning difficulty, has a blind spot: 10.3–22.9% of examples deemed impossible by sampling are actually solvable via activation grafting. The finding challenges current practices in RL training, data curation, and verifier design.
CREDENCE Framework Improves Automated Fact-Checking with Semantic Metrics and Convergence Analysis
The CREDENCE framework addresses key shortcomings in automated fact-checking by replacing Jaccard overlap metrics with Semantic-F1, a cosine similarity measure that improves accuracy by 15-32 percentage points. It also provides formal convergence theorems for repair pipelines and benchmarks across social media, encyclopedic, and news domains.
Reinforcement Learning Foundation Models: Synthetic MDPs Could Bridge the Gap
The paper by Zighem, Abdelrahman, and Vie argues that reinforcement learning (RL) lacks a foundation model equivalent to those for language and vision. They propose using synthetic Markov Decision Processes (MDPs), which are as feasible to generate as synthetic tabular data, and demonstrate with a Graph Attention Network trained entirely on synthetic MDPs that achieves competitive results without task-specific tuning.
New Robust Q-Learning Algorithm Tackles Mean-Field Control Under Wasserstein Uncertainty
A new robust Q-learning algorithm for discrete-time mean-field control problems under Wasserstein uncertainty in the common noise law combines quantization-and-projection with a Wasserstein dual reformulation. The algorithm, detailed in an arXiv preprint by researchers Laurière, Mathieu, Neufeld, Ariel, Park, and Kyunghyun, establishes convergence with finite-time iteration bounds for both synchronous and asynchronous learning. Numerical experiments on systemic risk and epidemic models illustrate its robustness-performance tradeoff and convergence behavior.
SleepMaMi: A Universal AI Foundation Model That Integrates Macro and Micro Sleep Structures
Researchers introduce SleepMaMi, a sleep foundation model that captures both full-night macro-structures and fine-grained micro-structures from polysomnography data. Pre-trained on over 20,000 PSG recordings (158K hours), it uses a hierarchical dual-encoder with Demographic-Guided Contrastive Learning and hybrid Masked Autoencoder objectives. SleepMaMi outperforms or matches state-of-the-art foundation models across diverse downstream tasks, enabling label-efficient clinical sleep analysis.
Emyx: New AI Model Generates All-Atom Proteins Faster and More Efficiently
Researchers have developed Emyx, a 140M-parameter conditional flow matching model for all-atom protein generation. Despite being the smallest model, Emyx outperforms both Proteína-Complexa and RFdiffusion3 on the AME enzyme design benchmark across success rate, structural novelty, scaffold diversity, and geometric validity, while training in just 682 GPU-hours—roughly 4× less than RFdiffusion3.
Researchers Propose Feature Selection to Improve Neural Additive Model Efficiency and Interpretability
A research paper proposes adding feature selection mechanisms to Neural Additive Models (NAM) and Neural Basis Models (NBM) to reduce computational costs and enable handling of feature interactions in high-dimensional datasets. The method updates selection weights during training, achieving better or comparable performance to state-of-the-art GAMs.
New Research Reveals How Visual Tokens Evolve Inside Vision-Language Models
A new computer vision paper from arXiv investigates how visual tokens are integrated into large language models (LLMs) under two paradigms: in-context prompting and layer-wise injection. The authors find that visual tokens enter the LLM as 'disguised visual context' lacking linguistic structure, then evolve differently depending on the integration architecture. They show that attention allocation alone is insufficient, and performance depends on the quality of visual representations at each layer.
Sequential DPO Study Reveals Non-Uniform Forgetting Across Multiple Preference Objectives
A study by Bhandari et al. on sequential Direct Preference Optimization (DPO) finds that later training objectives do not uniformly degrade earlier preferences. Using Llama-3.1-8B-Instruct, the research reveals that forgetting patterns vary from stability to positive transfer depending on objective compatibility and signal strength, offering guidance for multi-objective AI alignment in enterprises.
Deep Reinforcement Learning Framework Promises More Believable Game Characters
A paper on arXiv argues that deep reinforcement learning can create more believable game characters by learning from interactions or player data. The authors propose a framework with requirements suited for game development and identify bottlenecks to accelerate adoption.
LoRDO Algorithm Cuts Communication by 10x for Distributed AI Model Training
LoRDO (Low-Rank Distributed Optimization) unifies low-rank optimization with infrequent synchronization to reduce communication overhead in distributed training of foundation models. According to an arXiv paper, it achieves near-parity with low-rank DDP at scales 125M–720M parameters while cutting communication by approximately 10x, and shows further gains in very low-memory settings.
FlowMaps: Modeling Long-Term Multimodal Object Dynamics with Flow Matching
FlowMaps, a latent flow matching model, predicts multimodal distributions of future object locations in 3D space by learning from past human interactions. Tested in over 600 episodes, it outperforms state-of-the-art approaches for dynamic Object Navigation tasks in simulated and real environments. The research, published on arXiv, has potential applications for robotics in changing environments.
New Training-Free Method Enables Robots to Follow Personalized Commands Like 'Bring My Cup'
Researchers propose Visual Attentive Prompting (VAP), a training-free perceptual adapter that enables vision-language-action models to follow personalized commands by using reference images as visual prompts. VAP outperforms generic policies and token-learning baselines on simulation and real-world benchmarks.
Can In-Context Learning Enable Efficient Data Exploration for Enterprise AI?
A research paper investigates whether in-context learning (ICL) can enable intrinsic curiosity—automated data selection—without costly gradient updates. The authors prove that in general Markov decision processes, ICL-based rewards cannot unbiasedly estimate learning progress, but in non-temporal settings like active learning, they succeed. Controlled experiments validate the theory.
New Causal Discovery Algorithms BRIDGE and SKFM Use Lie Bracket Geometry to Uncover Latent Confounders
A new arXiv paper introduces two causal discovery algorithms, BRIDGE and SKFM, that use Lie bracket geometry to infer latent confounders. Built on Kan-Do-Calculus, these methods collapse the super-exponential space of possible DAGs and offer a new paradigm for causal discovery.
LOKI Memory-Free Method Improves Lifelong Knowledge Editing in Language Models by 14%
Researchers introduce LOKI, a memory-free method for lifelong knowledge editing in language models. It uses dynamic layer selection via the Hilbert-Schmidt Independence Criterion and projects gradient updates onto the null-space of model weights, eliminating the need for previous knowledge access. Experiments show up to 14% improvement in average accuracy over existing approaches.
Bid Farewell to Seesaw: New Framework Boosts Long-Tail Recommendation Accuracy Without Sacrificing Diversity
A new research paper from arxiv introduces HID (Hybrid Intent-based Dual Constraint Framework), a plug-and-play solution for session-based recommendation that simultaneously improves recommendation accuracy and long-tail item performance, eliminating the traditional trade-off. The framework uses hybrid intent learning and intent constraint loss to filter session-irrelevant noise.
DiverseDistill: New Knowledge Distillation Method Recovers Over 70% of Performance Gap Using Teacher Committees
Researchers propose DiverseDistill, a knowledge distillation framework that combines a large foundation model with domain-specific experts as a diverse committee. The method recovers 73–114% of the teacher-student performance gap on recommendation and vision tasks while requiring no parameter updates or architectural changes.
MENTOR: Reinforcement Learning via Flexible Teacher-Optimized Rewards for Tool-Use Distillation
A new research paper introduces MENTOR, a reinforcement learning framework that uses flexible teacher-optimized rewards to distill tool-use capabilities from large language models into small models. The approach improves out-of-domain generalization compared to supervised fine-tuning and strict reinforcement learning baselines.
New Framework for Class-Incremental Motion Forecasting Enables Autonomous Vehicles to Adapt to Novel Objects
Researchers introduce class-incremental motion forecasting, a setting where autonomous vehicles learn new object classes over time. They propose the first end-to-end framework that adapts to novel classes while mitigating catastrophic forgetting, using pseudo-labels and open-vocabulary segmentation. Evaluations on nuScenes and Argoverse 2 show preserved performance on known classes and effective adaptation to new ones.
Mitigating Simplicity Bias in OOD Detection through Object Co-occurrence Analysis
Researchers propose an object-centric OOD detection framework that leverages object co-occurrence patterns to overcome simplicity bias, achieving competitive results on near-OOD and full-spectrum settings.
New AI Framework PEGE Boosts HIV Detection by 15.4% in Networked Testing
A new AI framework called Policy-Embedded Graph Expansion (PEGE), combined with Dynamics-Driven Branching (DDB), improves HIV detection by 15.4% in networked testing. Developed with WHO and University of Witwatersrand, the approach supports UN Sustainable Development Goal 3.3 by making testing more efficient on incrementally revealed disease networks.
Technology Yann LeCun's new AI startup AMI Labs raises $1bn to build flexible intelligence beyond LLMs
Yann LeCun, former Meta chief AI scientist, has founded AMI Labs to develop a new AI architecture called JEPA, which aims to overcome the limitations of large language models (LLMs) in understanding the physical world. The startup raised over $1bn in seed funding from Nvidia and Jeff Bezos' private investment fund, marking one of Europe's largest seed rounds.
Technology British Police Predictive AI Models Quietly Abandoned After Staff Lost Trust in Results
An investigation by WIRED and partner outlets reveals that Avon and Somerset Police built at least 23 predictive analytics models, including risk scores for burglary, court non-appearance, and domestic abuse. At least two models were quietly abandoned after staff decided they could no longer trust them, while over 36,000 performance scores showed genuinely poor predictive performance. The program, centered on the Think Family Database holding records on half a million people, operated with limited transparency, raising concerns about public trust and algorithmic accountability.
DRFLOW Benchmark Targets Personalized Workflow Prediction for Enterprise AI Agents
Researchers introduce DRFLOW, a benchmark for evaluating AI agents on predicting personalized workflows from heterogeneous sources. The benchmark contains 100 tasks across five domains with 1,246 workflow steps grounded in over 3,900 sources, and defines seven diagnostic metrics. A reference agent, DRFLOW-Agent, shows improvement over baselines but highlights significant remaining challenges.
Machine Learning-Based Traffic Microsimulation Improves Crash Frequency Prediction, Study Finds
Researchers at the University of Leeds conducted microsimulation at five signalised intersections comparing a standard rule-based model and a machine learning behavior model. The ML model generated conflicts that led to crash predictions matching real-world data, while the rule-based model failed without calibration. The study highlights the potential of ML for proactive traffic safety assessment.
Multi-Agent Reinforcement Learning Achieves Superhuman Racing with 50% Fewer Collisions
A new study demonstrates that multi-agent reinforcement learning (MARL) allows quadrotors to achieve superhuman racing performance. Agents trained via league-based self-play outperformed champion humans at over 22 m/s and cut collision rates by 50% versus single-agent baselines, suggesting a new path for safe autonomous systems in shared spaces.
ROSE Benchmark Reveals Perception-to-Action Gap in Multimodal AI Models
The ROSE benchmark measures how reliably multimodal large language models (MLLMs) convert visual evidence into context-appropriate actions. Testing nine recent models, researchers found performance drops of up to 44.5 percentage points from counting to region-conditioned action, while humans achieve 98.8% accuracy.
Information Lattice Learning Framework Enables Interpretable Structure Learning for Probabilistic Graphical Models
A new paper by Haizi Yu and Lav R. Varshney introduces Information Lattice Learning (ILL) as a method for interpretable structure learning of probabilistic graphical models. ILL learns rules by alternately projecting signals onto partition lattices and lifting selected rules back, with applications in constraint-based factor graphs and maximum entropy modeling.
FlowEdit: Associative Memory Framework Cuts TTS Pronunciation Errors by 92.7% Without Retraining
FlowEdit, a new lifelong adaptation framework for flow-matching text-to-speech systems, corrects pronunciation errors on out-of-vocabulary proper nouns without retraining. By storing corrections as latent edits in a Modern Hopfield Network, it achieves a 92.7% reduction in Phoneme Error Rate on 312 multilingual proper nouns while maintaining speech quality.
New Framework Verifies Safety of Multi-Agent AI Communication for Autonomous Logistics
A new framework uses decision tree distillation to formally verify learned communication policies in multi-agent systems, targeting safety-critical autonomous logistics operations. The approach achieves 97.9% fidelity to neural policies and verifies 18 temporal logic properties with 88.9% satisfaction, including collision probabilities below 1% thresholds.
Residual-Space Evolutionary Optimization via Flow-based Generative Models
A new framework called residual-space evolutionary optimization addresses the challenge of data editing with non-differentiable objectives in flow-based generative models. By operating in residual space, it separates local exploitation (self-pollination) from broader exploration (cross-pollination). The method was validated on the MorphoMNIST benchmark and crystal data, showing balanced target alignment, instance preservation, and diversity.
Reinforcement-Aware Knowledge Distillation Boosts LLM Reasoning Efficiency
Researchers propose RL-aware distillation (RLAD) to address distribution mismatch and objective interference in knowledge distillation for LLM reasoning. The method uses Trust Region Ratio Distillation (TRRD) to selectively imitate teacher policies during reinforcement learning. RLAD outperforms offline distillation, standard GRPO, and KL-based on-policy distillation across logic and math benchmarks.
Beyond Reasoning Gains: Mitigating General-Capability Forgetting in Large Reasoning Models
A new research paper from arXiv shows that reinforcement learning with verifiable rewards (RLVR) can cause large reasoning models to forget foundational capabilities like perception and faithfulness. The authors propose RECAP, a replay strategy with dynamic objective reweighting that preserves general knowledge while maintaining reasoning gains.
MEAL Benchmark Enables Continuous Multi-Agent RL Training on 100 Tasks in Hours Using GPU Acceleration
Researchers introduced MEAL (Multi-agent Environments for Adaptive Learning), the first benchmark for continual multi-agent reinforcement learning. Using JAX and GPU acceleration, MEAL enables training on sequences of 100 tasks in hours on a single GPU, revealing failure modes not apparent at smaller scales. This addresses the limitation of previous benchmarks that only considered 3-10 sequential tasks due to CPU constraints.
FreeStyle: Scalable Style-Content Dual-Reference Generation via Community LoRA Mining
FreeStyle is a scalable dual-reference generation framework that leverages community LoRAs as compositional anchors for style and content. It introduces a two-stage curriculum with attention-level enrichment and frequency-aware RoPE modulation to suppress leakage from style references. The framework is evaluated on a new benchmark covering style similarity, content preservation, and leakage rejection, achieving a strong balance among these objectives.
Fine-Tuning LLMs for Vulnerability Detection Fails to Improve Security Reasoning, Study Finds
A new study introduces CWE-Trace, a framework for evaluating LLM vulnerability detection using Linux kernel samples. It finds that fine-tuning and data contamination do not improve security reasoning; detection accuracy remains near chance, and models lack genuine comprehension.
Repurposing a Speech Classifier for Guided Diffusion-Based Speech Generation
Researchers Makarov and Gerkmann propose a method to repurpose a conventionally trained speech classifier as the backbone for diffusion-based speech generation. By attaching a lightweight subnetwork and training only that under a Denoising Score Matching objective, they achieve high-quality speech synthesis with reduced memory footprint and computational cost compared to traditional classifier guidance that requires two separately trained models.
Multi-View Decompilation Improves LLM-Based Malware Classification, Study Finds
A new study shows that large language models (LLMs) classify decompiled code more accurately when given outputs from multiple decompilers rather than one. Researchers used Ghidra and RetDec to decompile benign and malicious binaries, finding that the multi-view approach improves malicious-class F1, mainly by increasing recall. The work suggests a simple, training-free method to enhance LLM-based malware triage in enterprise security operations.
New AI Research Shows Vision-Language Models Think Better with Visual Grounding
Researchers introduce visually grounded thinking, a reasoning process that interleaves natural-language thoughts with explicit point or box groundings to image regions. The method, using a scalable synthesis pipeline and grounding-aware reinforcement learning, consistently improves performance of Gemma3-4B-IT on counting and spatial reasoning benchmarks, with the 4B model matching or surpassing the 27B variant.
StreamKL Delivers up to 43× Speedup in Memory-Efficient Attention Distillation
Researchers propose StreamKL, a fused GPU primitive for Kullback-Leibler divergence in attention distillation. It eliminates quadratic memory materialization, enabling up to 43× and 14× speedups in forward and backward passes, and reduces extra HBM footprint to O(1).
Tri-Info Method Predicts VLA Model Failures with 83% Accuracy Across Real-World Tasks, Researchers Report
Researchers propose Tri-Info, a method using information theory to detect failures in Vision-Language-Action (VLA) models. It matches top baselines in-domain and achieves 83% accuracy on real-world tasks, with interpretable diagnostics.
Unsupervised Algorithms Cut Annotation Time by 78% for Industrial Semantic Segmentation
Researchers have demonstrated that unsupervised computer vision algorithms can reduce the annotation time for semantic segmentation tasks in industrial materials science by 78%, from 170 hours to 37 hours. The team created the largest public steel microstructure segmentation dataset and a benchmark deep learning model, validated by field experts and deployed in an industrial setting.
SL-S4Wave: Self-Supervised Learning Framework Improves ECG and EEG Analysis with State Space Models
Researchers propose SL-S4Wave, a self-supervised learning framework combining contrastive learning with structured state space models (S4) to analyze long-sequence physiological waveforms. The model outperforms state-of-the-art baselines in arrhythmia detection and EEG tasks, demonstrates strong label efficiency, and generalizes to unseen arrhythmia types.
Vero: An Open RL Recipe for General Visual Reasoning — A Fully Open Vision-Language Model Family
A new research paper introduces Vero, a family of fully open vision-language models (VLMs) that use reinforcement learning (RL) to achieve strong general visual reasoning. The team constructed a 600K-sample dataset from 59 datasets and designed task-routed rewards. Vero variants outperformed their base models by 2.9-5.4 points on average across a 30-benchmark suite, and the best variant surpassed a stronger closed model by 3.8 points. All code, data, and models are released publicly.
DeFrame: New Technique Debiases LLMs Against Subtle Framing Effects
Researchers at KAIST have identified framing disparity as an underexplored source of hidden bias in large language models (LLMs). Their proposed DeFrame method encourages consistent responses across semantically equivalent prompts, reducing overall bias and improving robustness against framing effects. The work has implications for enterprise AI deployments where fairness across demographics is critical.
SafeSpec: New Framework Boosts LLM Safety Without Sacrificing Inference Speed
Researchers propose SafeSpec, a safety-aware speculative inference framework that attaches a latent safety head to jointly evaluate semantic validity and safety in a single forward pass. On Qwen3-32B, it reduces attack success rates by 15% while preserving a 2.06x inference speedup on benign workloads, addressing the fundamental incompatibility between existing safety methods and speculative decoding.
Stabilizing the Q-Gradient Field for Policy Smoothness in Actor-Critic Methods
A team of researchers has introduced PAVE (Policy-Aware Value-field Equalization), a critic-centric regularization framework that stabilizes the Q-gradient field in continuous actor-critic reinforcement learning. The method addresses erratic high-frequency oscillations in learned policies without modifying the actor, achieving smoothness comparable to policy-side regularization while maintaining task performance.