Artificial Intelligence
Artificial Intelligence news — Monday 19 October 2026
150 stories from 10 sources, every one in English. Each links to its publisher.
-
Microsoft could hold an advantage with Copilot for Legal
AI Business
While the vendor is not the first in the AI for legal market, the Microsoft environment is entrenched in many law firms.
-
How Boston is becoming a physical AI hub
AI Business
Robotics talent, customers and investment are concentrating in Boston, and what that means for the U.S. physical AI industry.
-
Robot company bets touch, sound can give bots an edge over vision AI
AI Business
Neura’s latest foundation model combines vision, touch and audio to improve robotic perception, but training it on enough real-world data could prove difficult.
-
Making the internet better for AI agents might make it worse for everyone else
MIT Technology Review Biotech
This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here . Until recently, you could divide internet users into just two categories: bots and people. Websites…
-
Google, AI performance self-developed framework ‘Hans2’ released
AI Times (Korea)
Translated from Korean
-
Tencent, an AI memory integration that stored every chat bot.. . Shares with Clad·ChatGPT
AI Times (Korea)
Translated from Korean
-
“We were looking at it wrong”: The explosive claim that’s upended the AI world
MIT Technology Review Biotech
“I think I have good news about the nature of neural networks.” With those simple words of optimism, Nick Cammarata, a researcher at the AI startup Goodfire, kicked off a 26,000-word blog post that has knocked the AI world on its head…
-
El Rorian, a visual abstract AI ‘Grimphs’ released...In some evaluations before Astra
AI Times (Korea)
Translated from Korean
-
Antropicy Staff 91% “2030 AI Discover Nobel Prize” Forecast... Public 39%
AI Times (Korea)
Translated from Korean
-
SEND BIRD, AI Agent for Upscale and Enterprise Cooperation
AI Times (Korea)
Translated from Korean
-
China CXMT, Black-Switch Forecast On AI Demand...Changing 6% Drop
AI Times (Korea)
Translated from Korean
-
Antropic, an agent negotiating a lawyer... “The rules that should be kept when you are in vain”
AI Times (Korea)
Translated from Korean
-
Disney, the most popular character dotAI staff.. . 3 years
AI Times (Korea)
Translated from Korean
-
Achievement of the first-class technical work in Antropice to accelerated AI
AI Times (Korea)
Translated from Korean
-
5 companies, such as PuriosAI·Upstay, and so on, “Sover AI Pool Housing Infrastructure” joint development
AI Times (Korea)
Translated from Korean
-
NAVER PAGO expands into foreign travel guides, which will bring AI to translation
AI Times (Korea)
Translated from Korean
-
Dencom·Seoul, AI voice assistant training cooperation
AI Times (Korea)
Translated from Korean
-
OpenAI, Sejong Lee, Sejong Lee
AI Times (Korea)
Translated from Korean
-
Welcome to Google Gemini 4 Argon
Towards Data Science
A million-token output window changes what you can hand a model in one go, and the benchmarks and the bill show who it actually suits. The post Welcome to Google Gemini 4 Argon appeared first on Towards Data Science .
-
Extreme Heat May Be Making India’s Air Pollution Problem Worse
Asian Scientist
Researchers found that concentrations of organic carbon particles can rise sharply during extreme heat, adding to the pollution already in the air.
-
"&tropicy "Payable 5.5" partners"...Development tool new model name Itara capture
AI Times (Korea)
Translated from Korean
-
Antropic ‘Top 5 Clients’ Reflets are also openAI...Clip Rod Dependence
AI Times (Korea)
Translated from Korean
-
OpenAI, ‘GPT-Live’ decision-making API connection...’
AI Times (Korea)
Translated from Korean
-
We are committed to providing the best solutions for our customers and customers.
AI Times (Korea)
Translated from Korean
-
Mickey Mouse, drawing AI solution for manufacturing AX demand
AI Times (Korea)
Translated from Korean
-
Can Conformal Prediction Take You from Accurate to Reliable Machine Learning?
Towards Data Science
ML model adoption doesn't stop with acceptance of accuracy; it starts there. The post Can Conformal Prediction Take You from Accurate to Reliable Machine Learning? appeared first on Towards Data Science .
-
Copyright (c) 2014. All Rights Reserved.
AI Times (Korea)
Translated from Korean
-
Kimchee Festival Opening on 23rd...Kimchi Cuisine·Experience·Performance
AI Times (Korea)
Translated from Korean
-
IPO in front of Ensuke...MS, UK data center contract withdrawal
AI Times (Korea)
Translated from Korean
-
Antropic ‘Call’ causes Instagram account stopping. .Meta "Oin with Bot Activity"
AI Times (Korea)
Translated from Korean
-
Antropic, a quarterly 5% public interest fund... AI Job Impact Promotion
AI Times (Korea)
Translated from Korean
-
OpenAI, Image Model ‘Orora’ Three... 'GPT-image-3'
AI Times (Korea)
Translated from Korean
-
Import AI 477: In which I tell the IMF the Singularity has begun
Import AI Newsletter
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now Here’s a speech I gave on Saturday the 17th of October at…
-
How Many EEG Channels Do You Actually Need?
Towards Data Science
Testing sensor reduction, feature ranking, and generalization to unseen participants The post How Many EEG Channels Do You Actually Need? appeared first on Towards Data Science .
-
The Download: preserving the Arctic permafrost, and playful AI agents
MIT Technology Review Biotech
This is today’s edition of The Download , our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Inside the quest to combat Arctic fires and save the planet’s frozen carbon —James Temple In July…
-
Advancing Coding in Low-Resource Programming Languages
OpenAI Blog
Learn how OpenAI is improving ChatGPT across low-resource programming languages, with new evaluations for software engineering, SQLite, and COBOL migration.
-
USA "India·Indonesia·Laos Solar Importance Damages in the country's industry"...Bangping·Industry Customs Division
AI Times (Korea)
Translated from Korean
-
Copyright (c) 2015 FF. All Rights Reserved.
AI Times (Korea)
Translated from Korean
-
Spanish vineyard, low-end vertical photovoltaic test..."consumer quality impact limit"
AI Times (Korea)
Translated from Korean
-
AI Harms and Their Technical Causes
Towards Data Science
How technical choices can lead to safety failures, and what to do differently The post AI Harms and Their Technical Causes appeared first on Towards Data Science .
-
Inside the quest to combat Arctic fires and save the planet’s frozen carbon
MIT Technology Review Biotech
In the middle of an especially warm July in 2021, a wildfire ignited in a thick forest of black spruce and peeling paper birch trees on the northern campus of the University of Alaska, Fairbanks. The blaze, possibly sparked by lightning…
-
AI researchers stranded Claude on a virtual desert island. It started to “play.”
MIT Technology Review Biotech
What do AI agents do if you don’t tell them to do anything? The answer looks a lot like play. Researchers repeatedly stranded them alone on a virtual desert island for 30 hours in an experiment reminiscent of the movies Cast Away or Home…
-
[AIReader's Book] Deep Learning starting from the bottom 6
AI Times (Korea)
Translated from Korean
-
How far can we push computing?
MIT Technology Review Biotech
MIT Technology Review Explains: Let our writers untangle the complex, messy world of science and technology to help you understand what’s coming next. You can read more from the series here . On the evening of December 29, 1959, at an…
-
Learning to Correct Geometry in Generated Videos
arXiv cs.CV
arXiv:2610.23856v1 Announce Type: new Abstract: Generated videos often lose geometric consistency under camera motion: structures warp under viewpoint change, parallax becomes inconsistent, and smooth camera movement may collapse into…
-
Beyond Multi-Stage Pipelines: A Label-Correlation Ensemble for Multi-Label Retinal Fundus Disease Classification
arXiv cs.CV
arXiv:2610.23747v1 Announce Type: new Abstract: Retinal fundus images frequently carry more than one disease, and public fundus datasets are heterogeneous and severely long-tailed. Clinical triage suggests a staged design, in which a…
-
GARG-AML against Smurfing: A Scalable and Interpretable Graph-Based Framework for Anti-Money Laundering
arXiv cs.LG
arXiv:2506.04292v4 Announce Type: replace-cross Abstract: Money laundering, a global problem estimated to account for 2%-5% of the global GDP, required regulators to enforce stringent rules on financial institutions to combat it. One…
-
FlexPart: Text-Controlled Semantics-Aware 3D Part Segmentation at Flexible Granularity
arXiv cs.CV
arXiv:2610.23659v1 Announce Type: new Abstract: A 3D asset admits multiple valid part decompositions, while semantics-aware segmentation should reflect both the desired granularity and the parts a user wants to extract. Ideally, a…
-
GeoReason: Bridging Logical Reasoning and Spatial Fidelity in Remote Sensing Segmentation
arXiv cs.CV
arXiv:2610.23416v1 Announce Type: new Abstract: Multimodal Large Language Models (MLLMs) have significantly advanced instruction-driven perception through the "Embedding-as-Mask" paradigm. However, extending this capability to Remote…
-
Router Drift Control: Mitigating Off-Policy Degradation in Mixture-of-Experts Reinforcement Learning
arXiv cs.LG
arXiv:2610.21682v2 Announce Type: replace Abstract: Reinforcement learning (RL) can become off-policy due to training-inference mismatch, mini-batch updates, and asynchronous training. The resulting mismatch is especially harmful in…
-
LED: Improving Masked Diffusion Language Models by Lookahead Editing
arXiv cs.LG
arXiv:2610.23122v1 Announce Type: new Abstract: Diffusion large language models (dLLMs) have emerged as a promising alternative to autoregressive language models, enabling non-autoregressive text generation through parallel decoding…
-
LoongSage: A Production-Grade, High-Efficiency Agentic RL Framework for Frontier LLMs with Validated Recipes
arXiv cs.CL
arXiv:2610.16352v2 Announce Type: replace-cross Abstract: Reinforcement learning (RL) has emerged as a critical phase in the post-training pipeline of large language models (LLMs). As the frontier of LLM capabilities shifts from static…
-
FSPO: Policy-Consistent Risk and Pareto-Feasible Control for Budgeted LLM RL Post-Training
arXiv cs.CL
arXiv:2610.02828v2 Announce Type: replace-cross Abstract: Adaptive LLM reinforcement-learning post-training changes multiple training actuators online, including rollout temperature, group size, clipping, KL regularization, verifier…
-
Frame-Consistent Feed-Forward 3D Gaussian Splatting from Sparse Unposed Images
arXiv cs.CV
arXiv:2610.23039v1 Announce Type: new Abstract: Feed-forward Gaussian models have made novel view synthesis almost instantaneous. Feed-forward methods predict Gaussians directly, either in two stages or in a single merged pass. Using two…
-
Trust but Verify: Designing Region-Inhomogeneous Forward Dynamics for Discrete Diffusion
arXiv cs.LG
arXiv:2610.22380v1 Announce Type: new Abstract: Continuous diffusion treats the forward process as a design space; discrete language diffusion often uses absorbing corruption, leaving context fully trusted throughout generation. This makes…
-
Binding-Specific Relearning Savings for Test-Time Continual Learning
arXiv cs.LG
arXiv:2610.22381v1 Announce Type: new Abstract: When a fact becomes harder to elicit from a language model, has the model lost it, or has it merely become less accessible? Parameter-updating test-time continual learning agents are usually…
-
CF-NODE: Counterfactual Spatial Node Intervention for Test-Time Adaptation Under Domain Shift, Evaluated Across Five Diabetic Retinopathy Datasets
arXiv cs.LG
arXiv:2610.22437v1 Announce Type: new Abstract: Automated fundus-based diabetic retinopathy (DR) grading can substantially expand screening capacity, but source-trained models degrade sharply across different cameras, clinics, and patient…
-
Sparse Autoencoders Can Induce Local Geometry for Mitigating Catastrophic Forgetting
arXiv cs.LG
arXiv:2610.22439v1 Announce Type: new Abstract: Catastrophic forgetting remains a challenge in supervised continual learning, as adapting to new tasks causes representation drift and loss of previously learned information. This raises the…
-
Generative Value Classifier
arXiv cs.LG
arXiv:2610.22497v1 Announce Type: new Abstract: Q-learning is premised on learning a function to predict the value of a state-action pair, with the goal of using such a function to guide policy improvement. Most approaches learn this…
-
On the SPOT: Enhancing Speculative Decoding for Diffusion Language Models via Commit-Order Prediction
arXiv cs.LG
arXiv:2610.22543v1 Announce Type: new Abstract: Diffusion language models (DLMs) promise parallel text generation, but in practice face an unresolved speed-quality trade-off. Exact decoding fills in only one masked position per model pass…
-
Noise-shaped Sampling and Annealing for Scalable On-the-edge Training of Binary and Ternary AI Models
arXiv cs.LG
arXiv:2610.22621v1 Announce Type: new Abstract: The training loss of binary and ternary neural networks, expanded as a multilinear polynomial over the discrete weights, is structurally equivalent to the energy of a hierarchical high-order…
-
Edit Discovery in Human-AI Interactions
arXiv cs.LG
arXiv:2610.22661v1 Announce Type: new Abstract: Human-AI interactions are becoming a rich source of data for improving AI systems: while many model responses are already satisfactory, others expose failures that could become valuable…
-
Group-Graph Policy Optimization for Long-Horizon Agentic Reinforcement Learning
arXiv cs.CL
arXiv:2606.22995v2 Announce Type: replace-cross Abstract: Group-based Reinforcement Learning (RL) has significantly enhanced Large Language Models (LLMs) in agentic scenarios. To achieve finer-grained policy updates, recent agentic RL…
-
FrenchNews-7: Benchmarking Cross-Publisher French News Editorial Desk Classification
arXiv cs.CL
arXiv:2608.18097v2 Announce Type: replace Abstract: We present FrenchNews-7, a cross-publisher France-based French-language news editorial desk classification benchmark combining a large multi-outlet corpus, a URL-derived seven-class…
-
How Murky Before a VLM Knows It Cannot See? Accuracy, Confidence and Abstention on a Physical Visibility Axis
arXiv cs.CV
arXiv:2610.22981v1 Announce Type: new Abstract: Robustness studies of vision-language models (VLMs) grade degradation on nominal severity scales ("level 1-5") that carry no physical units, and they rarely ask whether a model's confidence…
-
Learning reveals invisible structure in low-rank RNNs
arXiv cs.LG
arXiv:2605.04115v2 Announce Type: replace Abstract: Learning in neural systems arises from synaptic changes that reshape the representations underlying behavior. While low-rank recurrent neural networks (RNNs) have emerged as a powerful…
-
DominoTree: Conditional Tree-Structured Drafting with Domino for Speculative Decoding
arXiv cs.CL
arXiv:2607.08642v4 Announce Type: replace Abstract: Speculative decoding speeds up large language model inference: a small drafter proposes several tokens, and the target model verifies them in one parallel pass. Drafting a tree lets the…
-
Pro-AI Bias in Large Language Models
arXiv cs.CL
arXiv:2601.13749v2 Announce Type: replace Abstract: Large language models (LLMs) are increasingly employed for decision-support across multiple domains. We ask whether these models favor artificial intelligence (AI) itself. We examine this…
-
EgoP$^{2}$C: Internalizing Plan-Perceive-Compose for Spatial Reasoning in Egocentric Video
arXiv cs.CV
arXiv:2610.22681v1 Announce Type: new Abstract: Answering complex spatial reasoning questions from egocentric video requires determining what evidence is needed, acquiring targeted observations, and combining them through multiple…
-
PresenceBench: An Exploratory Evaluation of Large Language Model Adaptation to User State
arXiv cs.CL
arXiv:2610.23625v1 Announce Type: cross Abstract: Presence-aware intelligence is the capacity of an AI system's intelligence layer to adapt appropriately to the functional state of the person it serves (here readiness, flow, and stability)…
-
The Statistical Fairness-Accuracy Frontier
arXiv cs.LG
arXiv:2508.17622v4 Announce Type: replace-cross Abstract: We study fairness-accuracy tradeoffs when a single predictive model must serve multiple demographic groups. A useful tool for understanding this tradeoff is the fairness-accuracy…
-
"You" Is Not Toxic: How LLM Labelers Rewrite the Web's Conversation Laws
arXiv cs.CL
arXiv:2610.23446v1 Announce Type: cross Abstract: What the Web community believes about toxic conversations, such as that toxicity rises in longer and deeper threads, was measured mostly with Google's Perspective API, which is shutting…
-
Three LiDARs, Twenty-Five Datasets: Cross-Source Generalization in Tree-Species Recognition
arXiv cs.CV
arXiv:2610.22606v1 Announce Type: new Abstract: FOR-species20K is the largest public LiDAR benchmark for tree-species classification, pooling 25 independently collected terrestrial, mobile, and airborne laser-scanning datasets across 33…
-
Adaptive Conditional Residual Sampling for Spectral Risk Optimisation under Decision-Dependent Uncertainty
arXiv cs.LG
arXiv:2603.12507v3 Announce Type: replace Abstract: We consider minimising a spectral risk objective, the expected cost plus a weighted conditional value-at-risk, when the distribution of the uncertain parameters depends on the decision…
-
Coverage Cliffs in Off-Policy Quantile Certification
arXiv cs.LG
arXiv:2610.22565v1 Announce Type: new Abstract: We study when logged contextual-bandit data can certify an upper tail quantile for a policy selected from the same sample. We construct a PAC-Bayes Bernstein lower-confidence envelope for the…
-
Sample, Mix, or Verify? Allocating Test-Time Compute under Imperfect Verifiers
arXiv cs.CL
arXiv:2610.23255v1 Announce Type: cross Abstract: A common way to make language models and agents more accurate is to spend more compute at inference time: generate many candidate solutions and submit the one a verifier prefers. Success…
-
Veil-Then-Reveal: Protecting PII Privacy via Reversible Pseudo-Anonymization in Zero-Trust User--LLM Interaction
arXiv cs.CL
arXiv:2610.22800v1 Announce Type: cross Abstract: Large language models (LLMs) increasingly process sensitive documents and personal information, yet sending prompts containing personally identifiable information (PII) to remote providers…
-
TriSAR-3D and PaintFusion: A Novel Trimodal LiDAR-RGB-Thermal Dataset and Method for 3D Person Detection in Search and Rescue
arXiv cs.CV
arXiv:2610.22509v1 Announce Type: new Abstract: Reliable 3D person detection is important for rescue robotics to give first responders situational awareness. Yet most multimodal 3D detection research targets automotive scenarios in which…
-
Bridging the Language Gap in Scholarly Data II: Enhancing Gender Inference for Chinese Names
arXiv cs.CL
arXiv:2610.23734v1 Announce Type: new Abstract: Perceived gender inference from names is important for inclusive scientometrics research, yet existing methods perform poorly on non-Western scholars. We address this gap for Chinese names…
-
MuninAI: A Self-Hosted Agentic Framework for Independent Research Groups
arXiv cs.CL
arXiv:2610.23480v1 Announce Type: new Abstract: Research groups that want a language model grounded in their own literature face a choice between hosted services that receive their unpublished work and self-hosted deployments that keep the…
-
Eliciting Specifications: When Does Interactive Clarification Help?
arXiv cs.CL
arXiv:2610.22382v1 Announce Type: cross Abstract: LLMs can generate Lean specifications from task descriptions, but those descriptions are often underspecified or ambiguous and may omit verifier-relevant edge cases, yielding imperfect…
-
RCPN: towards robust pseudo-labels in unsupervised CLIP adaptation via consistent prototypes and neighborhoods
arXiv cs.LG
arXiv:2507.22075v2 Announce Type: replace Abstract: In unsupervised adaptation for vision-language models such as CLIP, pseudo-labels derived from zero-shot predictions often exhibit significant noise, particularly under domain shifts or…
-
SpecXMaster Technical Report
arXiv cs.LG
arXiv:2603.23101v5 Announce Type: replace Abstract: Intelligent spectroscopy serves as a pivotal element in AI-driven closed-loop scientific discovery, functioning as the critical bridge between matter structure and artificial…
-
AOI: Turning Failed Trajectories into Training Signals for Autonomous Cloud Diagnosis
arXiv cs.LG
arXiv:2603.03378v4 Announce Type: replace Abstract: Large language model (LLM) agents offer a promising data-driven approach to automating Site Reliability Engineering (SRE), yet their enterprise deployment is constrained by three…
-
MGE-PO: Entropy-Aware Credit Assignment for Mathematical Reasoning
arXiv cs.CL
arXiv:2610.23245v1 Announce Type: new Abstract: GRPO assigns the same group-relative advantage to equally rewarded responses within a question, leaving their generation uncertainty unused. We introduce MGE-PO, centered on four-quadrant…
-
Learning without relearning physical symmetry: an SO(3)-equivariant convolutional network (CNN) on the sphere
arXiv cs.LG
arXiv:2610.23821v1 Announce Type: cross Abstract: Many of the most important datasets in the physical sciences live natively on the sphere, and these data often carry no preferred position or orientation. Using methods that do not reflect…
-
Interpretable experiential learning based on state history and global feedback
arXiv cs.LG
arXiv:2605.00940v2 Announce Type: replace Abstract: A new interpretable experiential learning model based on state history and global feedback is presented. It is capable of learning a behavioral model represented by a transition graph…
-
Human-Centered Learning Mechanics: Observable Entropy Control under Protected-Risk Constraints
arXiv cs.LG
arXiv:2605.22940v2 Announce Type: replace Abstract: Entropy reduction is not a safety criterion: it can leave predictions unchanged or accompany increasing risk for a protected group. We study entropy-regulated learning under separately…
-
Group-Graph Policy Optimization for Long-Horizon Agentic Reinforcement Learning
arXiv cs.LG
arXiv:2606.22995v2 Announce Type: replace Abstract: Group-based Reinforcement Learning (RL) has significantly enhanced Large Language Models (LLMs) in agentic scenarios. To achieve finer-grained policy updates, recent agentic RL frameworks…
-
Transferability of Learned States in Neural PDE Solvers
arXiv cs.LG
arXiv:2610.10972v2 Announce Type: replace Abstract: Assessing useful reuse in neural PDE solvers is challenging: final accuracy can reflect source learning and target-time computation. Our reuse contract separates solution accuracy…
-
Guarded Block Transfers for Polytope Distance: Linear Convergence with Explicit Constants and Gap-Certified Screening for SVM Training
arXiv cs.LG
arXiv:2610.15829v2 Announce Type: replace Abstract: Hard-margin SVM training is the distance problem between two convex hulls, which the Triangle Algorithm (TA) solves with a certified duality gap. We propose \emph{guarded block…
-
LoongSage: A Production-Grade, High-Efficiency Agentic RL Framework for Frontier LLMs with Validated Recipes
arXiv cs.LG
arXiv:2610.16352v2 Announce Type: replace Abstract: Reinforcement learning (RL) has emerged as a critical phase in the post-training pipeline of large language models (LLMs). As the frontier of LLM capabilities shifts from static reasoning…
-
Fully sequential Gaussian process regression for preference learning
arXiv cs.LG
arXiv:2610.19391v2 Announce Type: replace Abstract: We study continuous global optimization problems in which the objective function cannot be directly evaluated; rather, we observe only which of two selected design points appears to…
-
Spectral Indistinguishability and Learning-Curve Identifiability in Quadratic Networks
arXiv cs.LG
arXiv:2610.20678v2 Announce Type: replace Abstract: Motivated by information-organization sculpting, we study spectral summaries and scalar learning curves in centered quadratic networks. We construct analytic positive input densities with…
-
Sampling the XY model with an attention-based neural network
arXiv cs.LG
arXiv:2610.23783v1 Announce Type: cross Abstract: We present a transformer-based variational autoregressive network for sampling lattice systems with continuous degrees of freedom and apply it to the two-dimensional XY model near the…
-
Board gender diversity and emissions performance: Insights from panel regressions, machine learning, and explainable AI
arXiv cs.LG
arXiv:2510.00244v3 Announce Type: replace-cross Abstract: With European Union initiatives mandating gender quotas on corporate boards, a key question arises: Is greater board gender diversity (BGD) associated with better emissions…
-
Language steering in latent space to mitigate unintended code-switching
arXiv cs.LG
arXiv:2510.13849v4 Announce Type: replace-cross Abstract: Multilingual Large Language Models (LLMs) often exhibit hallucinations such as unintended code-switching, reducing reliability in downstream tasks. We propose latent-space language…
-
ECPO: Evidence-Coupled Policy Optimization for Verifier-Audited Candidate Ranking
arXiv cs.LG
arXiv:2605.21993v2 Announce Type: replace-cross Abstract: Ranking systems used for decision support should expose evidence that can be audited independently of the model's ranking scores. We study a closed-roster setting in which a policy…
-
Cross-Dialect Generalization Without Retraining: Benchmarks and Evaluation of Schema-Derived Constrained Decoding for MLIR
arXiv cs.LG
arXiv:2607.18254v2 Announce Type: replace-cross Abstract: Multi-Level Intermediate Representation (MLIR) underlies modern ML compiler infrastructure (TensorFlow, JAX/StableHLO, PyTorch Inductor, IREE), yet appears only in trace amounts in…
-
Capability-Gated Language Models: Security Composes, Utility Does Not
arXiv cs.LG
arXiv:2609.00445v3 Announce Type: replace-cross Abstract: Deployed language model safeguards (safety fine-tuning, filtering, unlearning) vary by principal only outside the model weights: filters are reconfigured, tiers are multiplied, and…
-
GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed
arXiv cs.LG
arXiv:2609.39600v2 Announce Type: replace-cross Abstract: Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens. We view grounding as visual evidence…
-
FSPO: Policy-Consistent Risk and Pareto-Feasible Control for Budgeted LLM RL Post-Training
arXiv cs.LG
arXiv:2610.02828v2 Announce Type: replace-cross Abstract: Adaptive LLM reinforcement-learning post-training changes multiple training actuators online, including rollout temperature, group size, clipping, KL regularization, verifier…
-
RealtimeWAM: One-Step Asynchronous World Action Models
arXiv cs.LG
arXiv:2610.06617v2 Announce Type: replace-cross Abstract: World Action Models (WAMs) incorporate visual representations from video generation backbones to guide action prediction. Recent efficient WAMs adopt Mixture-of-Transformers (MoT)…
-
Dependency Reasoning for Aspect Sentiment Quadruple Prediction
arXiv cs.LG
arXiv:2610.14873v2 Announce Type: replace-cross Abstract: Aspect Sentiment Quadruple Prediction (ASQP) is fundamentally a relational reasoning problem: identifying aspect-opinion-category-sentiment quadruples requires modeling structured…
-
Hypothesis Selection via Weighted Medians
arXiv cs.LG
arXiv:2610.20375v2 Announce Type: replace-cross Abstract: Given $n$ distributions (hypotheses) $\mathcal H=\{H_1,\ldots,H_n\}$ and samples from an unknown target distribution $P$, proper hypothesis selection seeks a distribution whose…
-
SafeAdapt-Poker: Counterfactual Risk Calibration for Budgeted Adaptation under Opponent Shift
arXiv cs.LG
arXiv:2610.20473v2 Announce Type: replace-cross Abstract: Opponent histories can mislead after behavioral shifts. SafeAdapt-Poker formulates budgeted adaptation under opponent shift as a dual-counterfactual sequential-control problem using…
-
FALCON: FPGA-Accelerated Latency-Certifiable Occupancy Networks for Camera-Based Autonomous Driving
arXiv cs.CV
arXiv:2610.22143v2 Announce Type: replace Abstract: A perception system that must brake for a pedestrian does not need a low average latency. It must deliver an answer before the deadline on every frame, the worst case included. Commodity…
-
Robust Function Calling for LLM Agents Under Interface Drift
arXiv cs.CL
arXiv:2610.23342v1 Announce Type: cross Abstract: Tool-augmented language-model agents rely on interface text: tool descriptions, argument schemas, and the candidate tool list shown at inference time. These interfaces often change after…
-
How Many, but How Sure? Class Competition as a Confidence Proxy in Multi-Class Density-Based Counting
arXiv cs.LG
arXiv:2610.23611v1 Announce Type: new Abstract: Multi-class density estimators can count objects in crowded scenes, but their class-wise outputs offer no indication of how reliably density mass has been allocated. We introduce a…
-
Practical and Private Hybrid ML Inference with Fully Homomorphic Encryption
arXiv cs.LG
arXiv:2509.01253v2 Announce Type: replace-cross Abstract: In contemporary cloud-based services, protecting users' sensitive data and ensuring the confidentiality of the server's model are critical. Fully homomorphic encryption (FHE)…
-
ActiveEvidence: Decision-Focused Evidence Graphs with Verifiable Acquisition Gaps
arXiv cs.LG
arXiv:2610.20532v2 Announce Type: replace-cross Abstract: A multimodal agent must decide whether visual evidence is worth acquiring, which candidate or resolution to query, and when further interaction can no longer change its decision. We…
-
Thinking from Above: Efficient Spatial Reasoning with Bird's-Eye-View Visual Tokens
arXiv cs.CV
arXiv:2610.22700v1 Announce Type: new Abstract: Existing video multimodal large language models (MLLMs) organize observations primarily in time, but embodied spatial reasoning requires organizing them in space. Consequently, their visual…
-
A Plausible Explanation Is Enough to Mislead Multimodal Judges
arXiv cs.CL
arXiv:2610.22631v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) increasingly serve as judges in training and evaluation pipelines, often assessing answers together with explanations. Although AI judges are…
-
MLLM Frame Subset Ensembling for Audio-Visual Video QA and MLLM-based Reranking for Ad-hoc Video Search in TRECVID 2025
arXiv cs.CV
arXiv:2610.23214v1 Announce Type: new Abstract: This paper presents the overview of the runs related to the Ad-hoc Video Search (AVS) and Video Question Answering (VQA) tracks of TRECVID 2025 on behalf of the CERTH-ITI team. For the AVS…
-
FlashControl: One-Step Controllable Generator via Distillation-Friendly Single-Stream Teachers
arXiv cs.CV
arXiv:2610.22537v1 Announce Type: new Abstract: Diffusion models have become a leading approach in visual generative modeling, particularly effective in text-to-image and controllable generation. Their main drawback is slow inference as it…
-
Deep Learning for Deepfake Image Detection: Comparative Model Evaluation and Gradient-Based Visual Attribution
arXiv cs.CV
arXiv:2610.22676v1 Announce Type: new Abstract: Synthetic facial imagery has made the distinction between authentic and generated content increasingly important for digital forensics and online trust. This study examines an explainable…
-
TransNSNet: A Transformer-Enhanced Multi-Scale Adversarial Framework for CT-to-PET Image Synthesis
arXiv cs.CV
arXiv:2610.23325v1 Announce Type: cross Abstract: This study investigates CT-to-PET image synthesis using TransNSNet, a Transformer-enhanced multi-scale adversarial framework. A paired CT-PET dataset was curated from the public non-small…
-
Self-Reflective Chain-of-Thought (SRCoT): Evidence-Based Text Generation via a Closed-Loop Retrieve--Reflect--Revise Pipeline on Edge LLMs
arXiv cs.CL
arXiv:2610.23311v1 Announce Type: cross Abstract: Large language models (LLMs) can generate fluent outputs that lack source support or conflict with local operational constraints. We introduce Self-Reflective Chain-of-Thought (SRCoT), a…
-
Cross-Layer Attention Probing for Fine-Grained Hallucination Detection
arXiv cs.CL
arXiv:2509.09700v2 Announce Type: replace Abstract: With the large-scale adoption of Large Language Models (LLMs) in various applications, there is a growing reliability concern due to their tendency to generate inaccurate text, i.e…
-
Self- and Other-Labels Induce Bidirectional Bias in LLM Judges
arXiv cs.CL
arXiv:2608.18091v3 Announce Type: replace Abstract: As LLM-as-a-judge becomes increasingly widespread, self-preference -- the tendency of a judge to favor its own outputs -- raises growing concerns about evaluation reliability. However…
-
Harmful YouTube Video Detection: A Taxonomy of Online Harm and MLLMs as Alternative Annotators
arXiv cs.CV
arXiv:2411.05854v2 Announce Type: replace-cross Abstract: Video platforms, used by billions globally, are criticized for exposing users to harmful content. Yet, detecting online harm remains challenging due to inconsistent approaches to…
-
Less is More: Efficient Black-box Attribution via Minimal Interpretable Subset Selection
arXiv cs.CV
arXiv:2504.00470v3 Announce Type: replace-cross Abstract: To develop a trustworthy AI system, it is essential to understand its behavior using attribution methods, which aim to identify the input regions that most influence the model's…
-
TV-AudioRemover: Joint Text-Visual Guided Sound Removal with Multi-Task Hard-Mixture Curriculum
arXiv cs.CV
arXiv:2609.25864v2 Announce Type: replace-cross Abstract: Visual object removal can eliminate a target from video frames, yet its acoustic trace persists in the soundtrack, causing obvious audio-visual inconsistency. Existing video…
-
MultiWorldBench: Do Independently Controlled Views Describe One Shared World?
arXiv cs.CV
arXiv:2610.11723v3 Announce Type: replace-cross Abstract: Multiplayer world models must ensure that independently controlled views remain consistent with a single shared and persistent world. We introduce MultiWorldBench, a diagnostic…
-
Riding the Fixed-Point Curve: Image Classification in the Continuous Thought Machine
arXiv cs.CV
arXiv:2610.19220v2 Announce Type: replace-cross Abstract: The Continuous Thought Machine (CTM) is a recurrent network that attends to inputs and makes predictions based on the synchronization between its neurons. On tasks spanning many…
-
Self-Evolving Coding Agents via Harness-Policy Optimization
arXiv cs.CL
arXiv:2610.22791v1 Announce Type: new Abstract: Large language model (LLM) agents have shown promising potential in coding tasks, and their performance is jointly determined by the policy model and the harness that mediates their…
-
HeRE: A Novel Benchmark for Hebrew Relation Extraction via LLM-Guided Denoising of Knowledge Graph Alignments
arXiv cs.CL
arXiv:2610.22542v1 Announce Type: new Abstract: While LLMs demonstrate strong capabilities in English Relation Extraction (RE) tasks, a significant gap remains in assessing their RE performance in low-resource languages, due to a lack of…
-
RAW-Explainer: Post-hoc Explanations of Path Reasoning Graph Neural Networks for Knowledge Graph Completion
arXiv cs.LG
arXiv:2506.12558v2 Announce Type: replace Abstract: Path reasoning graph neural networks have demonstrated state-of-the-art performance on knowledge graph completion tasks such as link prediction. However, interpreting their predictions…
-
HardEval: An Execution-Verified Benchmark for AI-Driven Security Hardening and Compliance Remediation
arXiv cs.LG
arXiv:2610.23685v1 Announce Type: cross Abstract: Implementing Security Technical Implementation Guide (STIG) controls remains a demanding operating-system administration task. Each natural-language security requirement must be translated…
-
The Optimal Ranking Tradeoffs Between Two Ranking Scores in Two-Class Classification
arXiv cs.LG
arXiv:2610.23294v1 Announce Type: cross Abstract: The accuracy, true negative rate (negative class recall), true positive rate (positive class recall), negative predictive value (negative class precision), positive predictive value…
-
Pearl-Chat: Engineering a native Luganda language model from first principles in JAX and Flax NNX
arXiv cs.CL
arXiv:2610.22529v1 Announce Type: new Abstract: Autoregressive foundation language models predominantly target high-resource Indo-European languages, imposing computational and morphological penalties on under-represented African…
-
Proportionality in Ranking Compression
arXiv cs.LG
arXiv:2610.23193v1 Announce Type: cross Abstract: Motivated by applications in reinforcement learning from human feedback, we consider a setting where a given collection of rankings needs to be compressed into a smaller collection of…
-
How Murky Before a VLM Knows It Cannot See? Accuracy, Confidence and Abstention on a Physical Visibility Axis
arXiv cs.LG
arXiv:2610.22981v1 Announce Type: cross Abstract: Robustness studies of vision-language models (VLMs) grade degradation on nominal severity scales ("level 1-5") that carry no physical units, and they rarely ask whether a model's confidence…
-
ActiveEvidence: Decision-Focused Evidence Graphs with Verifiable Acquisition Gaps
arXiv cs.CV
arXiv:2610.20532v2 Announce Type: replace-cross Abstract: A multimodal agent must decide whether visual evidence is worth acquiring, which candidate or resolution to query, and when further interaction can no longer change its decision. We…
-
Competence-Based Evaluation of Logical Reasoning in Large Language Models
arXiv cs.LG
arXiv:2610.22894v1 Announce Type: cross Abstract: Recent advances in large language models (LLMs) have brought impressive performance across diverse tasks; however, shortcomings remain in logical reasoning, including inconsistencies under…
-
Don't Retrain, Remix: Adaptation of Web Agents using Mixture Hypernetworks
arXiv cs.LG
arXiv:2610.22539v1 Announce Type: new Abstract: The web is not static or uniform. Websites that serve the same purpose can look and navigate very differently. Even within a single site, successive redesigns can change how familiar features…
-
Anytime Contracts for AI Services
arXiv cs.LG
arXiv:2610.22781v1 Announce Type: cross Abstract: AI services are prone to moral hazard, as model providers can silently route requests to cheaper, lower-quality models, and the inherent stochasticity of model outputs makes this…
-
One Communicator, Three Vendors: Extending the MoA Predictive Theory to a Real NVIDIA+AMD+CPU MPI Ensemble
arXiv cs.LG
arXiv:2610.22769v1 Announce Type: cross Abstract: The companion MPI-extension paper validated the first communication-bearing machine-shape ensemble: four NVIDIA A100 ranks and four CPU ranks on one NCSA Delta node under one Open MPI job…
-
Bernoulli Flow Models: Self-Consistent Generative Modeling for Binary Data
arXiv cs.CV
arXiv:2610.11362v2 Announce Type: replace-cross Abstract: Binary diffusion models typically require many function evaluations (NFEs) to generate high-quality samples, making inference costly. Reducing NFEs without sacrificing sample…
-
Shift-Reliance Alignment: Ranking Vision-Language Models by the Cues They Trust
arXiv cs.LG
arXiv:2610.22610v1 Announce Type: cross Abstract: Vision-language models (VLMs) are usually chosen from leaderboards, which give one ranking for every application. On questions about objects that a scene suggests but does not contain…
-
HeRE: A Novel Benchmark for Hebrew Relation Extraction via LLM-Guided Denoising of Knowledge Graph Alignments
arXiv cs.LG
arXiv:2610.22542v1 Announce Type: cross Abstract: While LLMs demonstrate strong capabilities in English Relation Extraction (RE) tasks, a significant gap remains in assessing their RE performance in low-resource languages, due to a lack of…
-
VideoHarness-SRSI: Structured Recursive Self-Improvement for Video Understanding Harness
arXiv cs.CV
arXiv:2610.19424v2 Announce Type: replace Abstract: LLM system capabilities depend not only on model parameters but also on the executable harnesses that organize information acquisition, tool use, reasoning, and memory. Recursive harness…
-
ROUTEAUDIT: Interaction-Aware Identification for Budgeted Multi-Verifier Routing
arXiv cs.LG
arXiv:2610.02808v2 Announce Type: replace-cross Abstract: Adaptive multi-verifier systems are commonly compared through endpoint quality-cost gaps, even when the verifier catalog, availability, accounting, information filtration, or scorer…
-
Scalable long-range machine learning interatomic potentials with electrical response for atomistic material modeling
arXiv cs.LG
arXiv:2610.23643v1 Announce Type: cross Abstract: Accurate atomistic simulations of materials under electric fields require long-range electrostatics and electrical-response modeling at length scales beyond the reach of conventional…
-
Has GPT-6-Astra Achieved Embodied Intelligence?
arXiv cs.LG
arXiv:2610.22471v1 Announce Type: cross Abstract: GPT-6-Astra demonstrates a level of competence in robotic manipulation and navigation that challenges the conventional paradigm of building specialized models for individual tasks and…
-
Adaptive conformal prediction on the Grassmannian
arXiv cs.LG
arXiv:2610.23508v1 Announce Type: cross Abstract: We introduce a conformal prediction framework for regression with Grassmann-valued responses. The proposed method constructs prediction regions in the form of metric balls directly on the…
-
Towards optimal selection schedules for online active learning
arXiv cs.LG
arXiv:2610.23477v1 Announce Type: cross Abstract: We consider the problem of online linear classification, where a stream of unlabeled samples is available, and annotation is costly. We investigate the optimal query policy, that prescribes…
-
An Adaptive-Horizon Graph Neural Operator for coupled surrogate morphodynamic modelling
arXiv cs.LG
arXiv:2610.22432v1 Announce Type: cross Abstract: Morphodynamic simulations require repeated coupling between hydrodynamic solvers, sediment transport modules, and bed updating. For long duration, multi-scenario simulations on large…
-
Towards Physically Grounded Articulated 3D Assets: A Multi-Agent Pipeline and Benchmark
arXiv cs.CV
arXiv:2610.13947v3 Announce Type: replace Abstract: In this paper, we study physically grounded articulated 3D asset generation---moving the goal from reproducing how objects look to capturing how they move and function. Inspired by the…
-
Characterising and Analysing Retrieval-Augmented Generation Techniques
arXiv cs.LG
arXiv:2610.23226v1 Announce Type: cross Abstract: The advent of Large Language Models (LLMs) across many application domains has increased the need for Retrieval-Augmented Generation (RAG) to reduce hallucinations and improve LLM…