Search: “routing”

From Detection to Refusal: Safer LLMs via Circuit-Guided Weight Scaling

arXiv:2609.00051v1 Announce Type: cross Abstract: Despite extensive alignment efforts, Large Language Models (LLMs) remain vulnerable to generating unsafe content under adversarial prompting, yet the

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Hardware

From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix

DGX agent

arXiv:2609.01572v1 Announce Type: new Abstract: Data-residency constraints force enterprises to self-host LLMs, but continuous adoption of newer models without decommissioning their predecessors expan

hardwarearxiv-cs-cl

2 Sept 2026

How Do Language Models Choose Between Context and Memory?

arXiv:2609.00753v1 Announce Type: cross Abstract: When contextual information conflicts with the knowledge stored in model parameters, activation directions can be used to decode and steer which sourc

Human-Anchored Factuality Evaluation with Strategic Annotation

arXiv:2609.00494v1 Announce Type: new Abstract: LLM-based factuality judges provide scalable evaluation signals, but their metrics are often systematically biased relative to human judgments. We study

safetyarxiv-cs-cl

MultiGait: A Multi-Sensor Multi-Perspective Multi-Session Biometric Inference Benchmark and its Dataset

arXiv:2609.01036v1 Announce Type: cross Abstract: A lack of suitable datasets has limited the research into the privacy risks of novel smart city sensors, such as thermal cameras, depth cameras, and l

Physically Plausible Video Generation via Visual-Semantic Chain-of-Events Conditioning

arXiv:2609.00656v1 Announce Type: new Abstract: Physically Plausible Video Generation (PPVG) seeks to synthesize videos consistent with physical principles, yet remains challenging due to underspecifi

SAM3-LoRA: Parameter-Efficient Adaptation of a Concept-Promptable Foundation Model for Multi-Class Structural Defect Segmentation

arXiv:2609.00469v1 Announce Type: new Abstract: Promptable segmentation foundation models such as SAM3 accept an open-vocabulary text concept and return every instance matching it, but adapting them t

ViPlan: A Benchmark for Visual Planning with Symbolic Predicates and Vision-Language Models

arXiv:2505.13180v3 Announce Type: replace Abstract: Integrating Large Language Models with symbolic planners is a promising direction for obtaining verifiable and grounded plans, with recent works ext

A Deep Latent Variable Framework for Jointly Modeling Missingness, Measurement Error, and Heterogeneity

arXiv:2608.30040v1 Announce Type: cross Abstract: Missing data, measurement error, and population heterogeneity are pervasive challenges in analyzing data arising from modern observational studies and

applicationsarxiv-cs-lg

Agents

Agentic AI uncovers conserved cross-tissue protein co-abundance programs inaccessible to single-dataset analysis

arXiv:2608.28990v1 Announce Type: new Abstract: Protein co-abundance clusters preserved across tissues can reveal shared disease mechanisms and candidate therapeutic targets, particularly when protein

agentsarxiv-cs-ai

Amortizing intractable inference in large language models

arXiv:2310.04363v3 Announce Type: replace-cross Abstract: Autoregressive large language models (LLMs) compress knowledge from their training data through next-token conditional distributions. This lim

safetyarxiv-cs-cl

An Agentic Retrobiosynthesis Framework with Learned Frontier Selection

arXiv:2608.30702v1 Announce Type: cross Abstract: Large language models are increasingly used as agents for multistep retrosynthesis, raising the question of how much their search policy contributes i

Are Online Skill and Memory Modules Always Worth Their Tokens? A Budget-Constrained Study of Web Agents

arXiv:2606.15017v2 Announce Type: replace Abstract: Online web agents often augment a base actor with memory, workflow, or skill modules. These modules can improve performance, but they also consume t

BLOOM-WILT: Logit Tilting for Behaviour Elicitation in Automated LLM Auditing

arXiv:2608.31105v1 Announce Type: new Abstract: Users of a deployed language model routinely encounter behaviours that testing almost never surfaces, since deployment puts the model through orders of

Budget-Aware Compression Pipeline for Single-GPU LLM Inference: Methods, Trade-offs, and Coupling Effects

arXiv:2608.30076v1 Announce Type: new Abstract: Single-GPU deployment of 70B-parameter language models on an NVIDIA GPU is constrained by device memory, long-context throughput, and engineering integr

Calibrating Small Language Models for Claim Check-Worthiness Detection

arXiv:2608.30731v1 Announce Type: cross Abstract: Assessing claim check-worthiness is an essential first step in automated fact-checking pipelines. This work is motivated by a real deployment challeng

applicationsarxiv-cs-ai

Can Video World Models Track Unobserved World States?

arXiv:2608.30692v1 Announce Type: new Abstract: Video world models are increasingly used as simulators, yet visual fidelity alone does not show that a model maintains the hidden state of the world. We

Defending Wearable VLMs Against Private Attribute Inference

arXiv:2608.28691v1 Announce Type: cross Abstract: Wearable VLM pipelines promise continuous multimodal assistance from egocentric visual capture: a user asks a task-driven question about the surroundi

During v3 testing for Fable 5.1 many of our requests were getting rejected due to Anthropic's safety guards misclassifying them as 'reverse …

safetyfrancois-chollet--x

During v3 testing for Fable 5.1 many of our requests were getting rejected due to Anthropic's safety guards misclassifying them as 'reverse engineering' We were in contact with the Anthropic team the

Financially Motivated Threat Actor BREEZE COMET Targets Brazil

applicationsgoogle-cloud-ai

Introduction Beginning in 2024 Mandiant investigated a string of compromises affecting Brazilian financial services, retail, and eCommerce organizations. Google Threat Intelligence Group (GTIG) tracks

GAFT: Geo-Anchored Fine-Tuning for Hazard Identification from Rare Failures

arXiv:2608.30858v1 Announce Type: cross Abstract: Off-road navigation can fail when physical structures induce irrecoverable states such as high-centering or entrapment, requiring human interventions.

Tutorials

Generative Retrieval for E-commerce: Jointly Learning Embedding and Codebook with Same Product Cluster

arXiv:2608.30606v1 Announce Type: cross Abstract: With the development of large language models (LLMs), generative retrieval is becoming increasingly important in e-commerce scenarios. Current mainstr

tutorialsarxiv-cs-ai

Keyframe-Centric State-Space Modeling for Burst Image Super-Resolution

arXiv:2503.19634v2 Announce Type: replace Abstract: Burst image super-resolution (BISR) reconstructs a high-resolution keyframe by aggregating complementary sub-pixel evidence from a short burst of lo

Learning to Restore More: Continual Capability Expansion for Pretrained Image Restoration Models

arXiv:2608.30305v1 Announce Type: new Abstract: Image restoration models are typically trained with a fixed set of capabilities. When new restoration requirements emerge, existing solutions usually tr

LLM Judges Verify Presence, Not Absence: Omission Blindness in AI Clinical Notes and What Recovers It

arXiv:2608.31016v1 Announce Type: cross Abstract: Ambient AI scribes draft clinical notes, and published audits find their dominant error is omission: information the encounter established that the no

Off-Manifold Refinement: Guiding Video Generators with a Frozen World Model

arXiv:2608.29904v1 Announce Type: new Abstract: Modern video generators routinely fail at physical dynamics: objects float, trajectories violate gravity, contacts vanish. Standard denoising and flow-m

On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability

arXiv:2608.30320v1 Announce Type: new Abstract: We describe the architecture and ablations of Qwen3.8-Flash-Next, a sparse mixture-of-experts model with 125B parameters, 6B activated per token, and ad

PAGE-RAG: Provenance-Aware Graph Evidence Promotion for Fixed-Budget Multi-hop Retrieval-Augmented Generation

arXiv:2608.29753v1 Announce Type: new Abstract: Multi-hop question answering in retrieval-augmented gener?ation (RAG) often benefits from retrieving beyond the few candidates that will finally be read

researcharxiv-cs-ai

PLC-DPO: Posterior Label Correction in Noisy and Ambiguous Preference Optimization

arXiv:2608.30597v1 Announce Type: cross Abstract: Direct Preference Optimization (DPO) simplifies alignment through pairwise comparisons but assumes all observed preferences are reliable. Real data of

Real-Time Deadlines Reveal Fragile Temporal Adaptation in LLM Strategic Dialogues

arXiv:2601.13206v2 Announce Type: replace Abstract: Large Language Models (LLMs) generate text token-by-token in discrete time, yet real-world communication, from therapy sessions to business negotiat

Recognition-Refusal Misalignment in LLMs: Why Models Answer Structurally Unanswerable Questions

arXiv:2608.29109v1 Announce Type: new Abstract: Large language models often answer structurally unanswerable questions, such as computing cot(-540{eg}) or evaluating (1).startswith('1'), instead of ab

safetyarxiv-cs-cl

Reference-Grafting Matches Fine-Tuning at Eliciting Sandbagged Capabilities

arXiv:2608.29458v1 Announce Type: cross Abstract: Sandbagging, in which a model deliberately underperforms on an evaluation despite retaining the underlying capability, threatens the safety evaluation

Agents

SkillZip Pro: Execution-Aware Dynamic Compression of Progressively Loaded Skills for Self-Evolving Agents

arXiv:2608.30785v1 Announce Type: new Abstract: Production agent skills are directory bundles, not isolated prompts. The root is loaded at activation; references, schemas, scripts, assets, and nested

agentsarxiv-cs-ai

STEP: A Modular Silent Trial Engine for Operational Evaluation of Digital Pathology AI in Routine Workflow

arXiv:2608.28708v1 Announce Type: cross Abstract: Prospective silent trials provide an important bridge between retrospective validation of artificial intelligence (AI) models and their use in clinica

local-aiarxiv-cs-cv

The Differential Reasoning Router: Operationalizing Cost-Aware LLM Annotation in E-commerce

arXiv:2608.30224v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly used to annotate structured product data in e-commerce, but early deployment often begins as a cold-star

applicationsarxiv-cs-ai

Token-Efficient Data Reasoning Agents via Adaptive Structuring of Unstructured Data

arXiv:2608.31082v1 Announce Type: new Abstract: Valuable data remains embedded in unstructured sources: web pages, reports, contracts, filings, earnings calls, and PDFs. The big bet in enterprise AI i

Uncertainty-Aware Multimodal Anti-UAV Detection via Evidential Fusion and Conflict-Discounted Belief Aggregation

arXiv:2608.29235v1 Announce Type: new Abstract: Anti-UAV perception systems must remain reliable when sensor streams degrade under occlusion, fast motion, or modality-specific failure. Existing multim

Unsupervised Multi-Scale Gromov-Wasserstein Hypergraph Alignment

arXiv:2608.29635v1 Announce Type: new Abstract: We study unsupervised hypergraph alignment, where the goal is to infer node correspondences between two hypergraphs using only structural information, w

safetyarxiv-cs-lg

Whole-Slide Image Analysis under Realistic Few-Shot Annotation Protocols

arXiv:2608.30420v1 Announce Type: cross Abstract: Automating the analysis of whole-slide images has high clinical value, since characterizing cancers requires examining them in detail. Such analysis i

local-aiarxiv-cs-ai

You Do Not Fully Utilize Transformer's Representation Capacity

arXiv:2502.09245v3 Announce Type: replace-cross Abstract: In contrast to RNNs, which compress their history into a single hidden state, Transformers can attend to all past tokens directly. However, st

researcharxiv-cs-cl

Zero-shot Generalizable Graph Anomaly Detection with Mixture of Riemannian Experts

arXiv:2602.06859v3 Announce Type: replace-cross Abstract: Graph Anomaly Detection (GAD) aims to identify irregular patterns in graph data, and recent works have explored zero-shot generalist GAD to en

Dynamic Alignment Compensation for Hallucination Mitigation in Large Vision-Language Models

arXiv:2608.28058v1 Announce Type: new Abstract: Large Vision-Language Models (LVLMs) remain prone to hallucinations, producing responses that are irrelevant or inconsistent with the multimodal input.

safetyarxiv-cs-cv

GeBDA: Building Damage Assessment as Text-Based Sequence Prediction

arXiv:2608.28567v1 Announce Type: new Abstract: Conventionally, Building Damage Assessment (BDA) is tackled either with dedicated network architectures or by fine-tuning geospatial image foundation mo

Ladders in Chaos: When, How, (and Perhaps Why) Does Test-Time Scaling Improve LLM Machine Translation

arXiv:2608.28496v1 Announce Type: new Abstract: Two forms of test-time scaling for Large Language Models (LLMs) have emerged as effective and widely adopted paradigms: sequential, in which later answe

researcharxiv-cs-cl

Mixture of Multicenter Experts in Multimodal AI for Debiased Radiotherapy Target Delineation

arXiv:2410.00046v4 Announce Type: replace-cross Abstract: Clinical decision-making reflects diverse strategies shaped by regional patient populations and institutional protocols. However, most existin

local-aiarxiv-cs-cv

Select, Label, Evaluate: Active Testing in NLP

arXiv:2603.21840v2 Announce Type: replace Abstract: Human annotation cost and time remain significant bottlenecks in Natural Language Processing (NLP), with test data annotation being particularly exp

researcharxiv-cs-cl

StreamEMS: Streaming Video Understanding with Self-Evolving Memory Scheme for Vision-Language Models

arXiv:2608.27881v1 Announce Type: new Abstract: Recently, many streaming video understanding methods have been proposed by constructing an external memory to store historical data for computational re

TACIT-Switch: Cost-Aware Model Escalation for LLM Agents from Censored Supervision

arXiv:2608.27911v1 Announce Type: new Abstract: Agents with smaller language-model backbones are less expensive but can drift into persistent failure modes, whereas those with larger backbones are gen

local-aiarxiv-cs-lg