Search: “llm”

LongVU-TTT: Causal Test-Time Training for Visual Resampling in Long Video Understanding

arXiv:2608.25729v1 Announce Type: new Abstract: Long-video MLLMs must model temporal change before a limited visual-token budget removes most frame evidence. We introduce LongVU-TTT, which inserts a c

researcharxiv-cs-cv

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Multi-Modal Anomaly Detection: A Survey

DGX agent

arXiv:2608.24937v1 Announce Type: new Abstract: Multi-Modal Anomaly Detection (MMAD) detects rare abnormal events from heterogeneous data sources and is increasingly used in safety- and reliability-cr

safetyarxiv-cs-lg

27 Aug 2026

MulVec: Fine-Grained Role-Aware Matching for Training-Free Zero-Shot Composed Image Retrieval

arXiv:2608.25305v1 Announce Type: new Abstract: Training-free zero-shot composed image retrieval finds a target image in a gallery from a reference image and a text edit without learning from task-spe

researcharxiv-cs-cv

Not All Attention Heads Contribute to Critical Visual Token Selection: Head-Aware Pruning Matters More

arXiv:2608.25332v1 Announce Type: new Abstract: Vision-Language Models (VLMs) have exhibited impressive performance across diverse visual scenarios. However, this success comes at the cost of explosiv

researcharxiv-cs-cv

One Form to Transfer Them All: Pretraining Multilingual Language Models Beyond Native Orthography

arXiv:2608.25904v1 Announce Type: new Abstract: Multilingual language models transfer knowledge across languages through shared subword vocabulary, a mechanism that breaks down when related languages

researcharxiv-cs-cl

One Symptom, Three Levers: A Critical Review of On-Policy Self-Distillation

arXiv:2608.25936v1 Announce Type: cross Abstract: On-policy distillation trains a language model on its own generations while a teacher scores them token by token. It combines the dense supervision of

safetyarxiv-cs-cl

Prefix Sliding for efficient test-time scaling

arXiv:2608.26070v1 Announce Type: new Abstract: Test-time scaling uses extra test-time compute to improve performance, such as letting language models reason longer when solving a problem. As models k

researcharxiv-cs-cl

Provenance Before Prose: Claim-Locked Reporting

arXiv:2608.25336v1 Announce Type: new Abstract: Large language models (LLMs) can fluently verbalize statistical evidence, yet statistical reports can still drift numerical values, invert effect direct

Query-Side Attacks on GNN-Based KGQA: Tracing Failures from Entity Linking to Answer Generation

arXiv:2608.25922v1 Announce Type: new Abstract: GNN-based Knowledge Graph Question Answering (KGQA) pipelines process queries through four discrete stages: entity linking, subgraph retrieval, GNN reas

Refusal geometry reflects refusal training: diverse refusal prefixes can raise stable rank and weaken refusal vector ablation attacks

arXiv:2608.25390v1 Announce Type: new Abstract: Refusal training protects AI models from jailbreaks by training models to decline unsafe queries, reducing the risk of misuse. Recent work finds that re

Resolving Multi-Modal Regression by Difference-Quotient-Based Clustering:Fast Coarse Conditional-Label Assignment

arXiv:2608.25467v1 Announce Type: new Abstract: Multimodal regression suffers from the mean-collapse pathology: under squared loss, an unconstrained regressor converges to the conditional mean, which

researcharxiv-cs-lg

SimLens for Early Exit in Large Language Models: Eliciting Accurate Latent Predictions with One More Token

arXiv:2507.17618v3 Announce Type: replace Abstract: Intermediate-layer predictions in large language models (LLMs) are informative but hard to decode accurately, especially at early layers. Existing l

Think-Probe-Respond: Improving Large Language Models as Judges of Research Idea Novelty

arXiv:2608.25660v1 Announce Type: new Abstract: Automated novelty judgment can accelerate scientific discovery by enabling efficient evaluation, refinement, and comparison of research ideas. While lar

safetyarxiv-cs-cl

Towards Reliable, Generalizable, and Specific In-Context Knowledge Editing via Multi-Objective Reinforcement Learning

model-releasesarxiv-cs-lg

arXiv:2608.25100v1 Announce Type: cross Abstract: Large Language Models (LLMs) are powerful but limited by static parametric knowledge that becomes outdated once pretraining ends. Knowledge editing ad

Unfolding Scientific Papers into Multi-Turn Generation Trajectories for Continued Pre-Training

arXiv:2608.25826v1 Announce Type: new Abstract: A recent line of synthetic-data work reconstructs the thinking behind existing text rather than rewriting the text itself, but it operates on short web

Virgil: Navigating Explainability for Transformer-based Language Models

arXiv:2608.25555v1 Announce Type: new Abstract: Explainability for transformer-based language models is becoming crucial as these systems are deployed in high-stakes applications. As a result, the eco

researcharxiv-cs-cl

When Stale Constraints Go Unchecked: Budgeted Verification Failures in Inherited Agent Memory

arXiv:2608.25553v1 Announce Type: cross Abstract: An agent that inherits a consolidated memory may inherit a constraint that was true when written and has since been withdrawn by a newer authoritative

safetyarxiv-cs-cl

Tutorials

A Behavior-Guided Online Probabilistic Forecasting Method for Electric vehicle Charging Loads

arXiv:2608.24441v1 Announce Type: new Abstract: Electric vehicle (EV) charging loads exhibit strong behavioral heterogeneity and temporal variability, posing significant challenges for online probabil

tutorialsarxiv-cs-ai

Are Android GUI Agents Robust Against Runtime Anomalies? AnTrap: Evaluating Agents in Dynamic Adversarial Environments

arXiv:2608.24099v1 Announce Type: new Abstract: GUI agents often encounter dynamic anomalies when deployed on Android devices, from unexpected pop-ups to action misuse, yet existing benchmarks lack sy

BenchBench-Protocol: Evaluating Real-World Wet-Lab Protocol Reasoning and Modification

arXiv:2608.23898v1 Announce Type: new Abstract: We introduce BenchBench-Protocol, a benchmark for large language models of 149 protocol-modification tasks recovered from modifications that scientists

CAFE: Self-Improving Search Agents Need Co-Evolving Feedback

arXiv:2608.24794v1 Announce Type: new Abstract: Outcome-supervised search agents learn when and how to retrieve evidence, but terminal rewards neither localize intermediate errors nor redirect an ongo

Can a Dynamic Internal Field Govern a Transformer's Cognition? Certifiability, not Superiority, in Homeostatic Compute Control

arXiv:2608.24319v1 Announce Type: new Abstract: An intelligent system does not merely reason: it governs its own reasoning - how much to compute, when to stop, which module to activate. Can that role

Causal Analysis for Time Series Foundation Models

arXiv:2608.24303v1 Announce Type: new Abstract: Transitioning from bespoke time series models towards time series foundation models changes the relationship of model and application from one-to-one to

Comparing Uncertainty Measurement and Mitigation Methods for Large Language Models: A Systematic Review

arXiv:2504.18346v4 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have been transformative across many domains. However, hallucination, i.e., confidently outputting incorrect info

Agents

Discovering Adaptive Transmission Programs for Collective Innovation

arXiv:2608.24545v1 Announce Type: new Abstract: Human collective intelligence depends on transmission processes: who shares what with whom, how, and when. While these processes emerge from individual

agentsarxiv-cs-ai

Effective Learning Rate Governs Loss Dynamics in Language Model Pretraining

model-releasesarxiv-cs-lg

arXiv:2608.24814v1 Announce Type: new Abstract: We uncover ELR collapse in language model pretraining: learning rate (LR) and parameter norm govern loss dynamics primarily through their ratio, the eff

EmoTra-TTS: Smooth Intra-Utterance Emotion Transitions for Speech Synthesis

arXiv:2608.23791v1 Announce Type: cross Abstract: Psychological research on emotion dynamics has established that human affect is a continuous, evolving process: emotions rise, decay, and transition w

Local Ai

exttt{findr}: Transparent and Fair Credit Risk Decisions through Semi-Structured Regressions

arXiv:2608.24582v1 Announce Type: cross Abstract: Credit risk models increasingly need to combine predictive accuracy with transparent explanations and auditable fairness constraints. Logistic regress

local-aiarxiv-cs-ai

Local Ai

FedV-KGQA: Multi-Hop Question Answering over Vertically Partitioned Knowledge Graphs

arXiv:2608.24846v1 Announce Type: new Abstract: Real-world data for knowledge graph question answering is often distributed across different organizations due to governance and data sovereignty constr

local-aiarxiv-cs-ai

Applications

FLARE: A Systematic, Uncertainty-Aware Framework for Evidence-Based Adoption of Artificial Intelligence in Healthcare

arXiv:2608.23643v1 Announce Type: new Abstract: Artificial intelligence is increasingly being introduced into healthcare workflows, yet most evaluations emphasize model accuracy rather than whether ad

applicationsarxiv-cs-ai

Applications

Generating Intervention Hypotheses using Explainable Explanations on Graphs: G2I, a Two-Stage Greedy Framework

arXiv:2608.23835v1 Announce Type: new Abstract: Real-world decision-making in public health and social science can greatly benefit from predictive models, yet translating predictions into effective in

applicationsarxiv-cs-lg

Granite.Trust Policy Tools: Shareable, Actionable Policies for Generative AI Applications

arXiv:2608.23870v1 Announce Type: new Abstract: When it comes to safety policies for generative AI, one size does not fit all. Each organization and use case needs to mitigate different risks dependin

safetyarxiv-cs-ai

How much of a measured AI preference is the model, and how much is the instrument?

arXiv:2608.23641v1 Announce Type: new Abstract: Model welfare research infers what a model prefers from the answers returned to prompts written to elicit preferences. Keeling et al. (2024), Mazeika et

Applications

LEMMA-RCA: A Large Multi-modal Multi-domain Dataset for Root Cause Analysis

arXiv:2406.05375v4 Announce Type: replace Abstract: Root cause analysis (RCA) is crucial for enhancing the reliability and performance of complex systems. However, progress in this field has been hind

applicationsarxiv-cs-ai

Local Ai

Maia 200: A Software Defined Dataflow System for Large-scale AI Acceleration

arXiv:2608.24664v1 Announce Type: cross Abstract: We introduce Maia 200, an advanced AI accelerator delivering high performance-10 145 Tflop/s FP4 and 5072 Tflop/s FP8 within a 750W TDP and 7 TB/s HBM

local-aiarxiv-cs-ai

Mitigating Exploration Bias in RL for Multi-Instruction Following

arXiv:2608.23830v1 Announce Type: cross Abstract: RL has emerged as a powerful paradigm for enhancing the instruction following capabilities of LLMs. While existing training recipes achieve substantia

MolGA: Molecular Graph Adaptation with Pre-trained 2D Graph Encoder

arXiv:2510.07289v2 Announce Type: replace Abstract: Molecular graph representation learning is widely used in chemical and biomedical research. While pre-trained 2D graph encoders have demonstrated st

PatientHub: A Unified Framework for Patient Simulation

arXiv:2602.11684v2 Announce Type: replace-cross Abstract: As Large Language Models increasingly power role-playing applications, simulating patients has become a valuable tool for training counselors

Persistent Cross Entropy

arXiv:2608.24549v1 Announce Type: new Abstract: Persistent entropy is the Shannon entropy of a persistence-based probability measure defined on a persistence diagram. However, its cross-entropy versio

researcharxiv-cs-lg

Agents

PinSieve: Production Selective VLM Serving and a Governed Memory Flywheel for Enterprise Content-Quality Triage

arXiv:2608.24040v1 Announce Type: new Abstract: Enterprise AI agents in production often need to be bounded, stateful, observable, and governable rather than fully autonomous. We present PinSieve, a p

agentsarxiv-cs-lg

QABBA: Error-Guaranteed Symbolic Time-Series Compression via Integer-Quantized Aggregation

model-releasesarxiv-cs-lg

arXiv:2411.15209v3 Announce Type: replace Abstract: The expansion of time-series data from sensors and monitoring systems has made compact representations increasingly important. Such representations

Recursive Agentic Reasoning

arXiv:2608.23956v1 Announce Type: new Abstract: Test-time reasoning methods such as iterative refinement, decomposition, and repeated sampling are often evaluated in isolation, making their gains diff

RePolicy: Reinforcement Learning for Safety-Policy Invocation in Agent Safeguards

arXiv:2608.24275v1 Announce Type: new Abstract: Safeguarding language model agents requires assessing complete execution trajectories under context-dependent safety policies. Existing policy-aware saf

safetyarxiv-cs-ai

ReproAgent: Contract-Guided Paper-to-Code Reproduction

arXiv:2608.24291v1 Announce Type: new Abstract: Paper-to-code reproduction asks scientific AI agents to turn research papers into executable repositories that preserve the paper's method, protocol and

Robust Code RL via Faulty-Code-Driven Test case Synthesis and Dense Reward Shaping

arXiv:2608.24135v1 Announce Type: new Abstract: Reinforcement learning from verifiable rewards (RLVR) has emerged as a pivotal technique for enhancing the code generation capabilities of Large Languag

SAGE: From Direct Answering to Evidence-Grounded Inference for Chinese Ancient Document Understanding

arXiv:2608.24011v1 Announce Type: cross Abstract: Chinese ancient document understanding demands complex visual, linguistic, and historical reasoning. Current Large Vision-Language Models (LVLMs) typi

Semantic Overlays: Mitigating Prompt Injection with Annotations Beyond Tokens and Steering Vectors

arXiv:2608.23873v1 Announce Type: new Abstract: Everything a language model sees is tokens. The serving stack knows what each span is -- user input, tool output, instructions -- but the model must kee

Hardware

ShardMeter: Sharded and Geo-Distributed Training Without the Guesswork

arXiv:2608.23840v1 Announce Type: cross Abstract: Training large-scale AI models often outgrows a single data center, demanding sharded, multi-cluster, and decentralized training. However, the huge sp

hardwarearxiv-cs-ai