Search: “evaluation”

Lost in Translation: Why Lower Bias Scores Don't Mean Fairer Models

arXiv:2610.20479v1 Announce Type: new Abstract: Bias audits of multilingual language models usually compare a model's average stereotype score in English with its average in another language and read

safetyarxiv-cs-cl

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

LURE: a Phase-Labelled Synthetic Corpus of Online-Grooming Exchanges

DGX agent

arXiv:2610.15908v1 Announce Type: new Abstract: Phase-annotated conversational data is essential for studying the progression of online grooming of minors and for developing early warning tools. Howev

model-releasesarxiv-cs-ai

15 Oct 2026

MakSafe: A Spatiotemporal Semantic Framework for Formal Specification and Compliance Checking of Road Vehicle Safety Rules

model-releasesarxiv-cs-cl

arXiv:2610.19735v1 Announce Type: new Abstract: Scenario-based validation is essential for automated road vehicles but cannot cover the full variability of openworld traffic. We present MakSafe, a for

MMLA-RSI: Recursive Self-Improvement through RTT and RTMA

arXiv:2610.20294v1 Announce Type: new Abstract: An agent can improve its behavior within an unresolved problem by changing how it acts and which experience it retains. We propose MMLA-RSI, a recursive

safetyarxiv-cs-cl

MonoMIP: Multi-Height Inverse Perspective Mapping for Monocular 3D Object Detection

arXiv:2610.19071v1 Announce Type: new Abstract: Monocular 3D object detection methods increasingly rely on depth-guided transformers, but they operate entirely in the perspective image plane where spa

researcharxiv-cs-cv

Noisy-Label Calibration with Multiple Unreliable Labels

arXiv:2610.19543v1 Announce Type: cross Abstract: In medical annotation and crowdsourcing, multiple annotators assign labels to each example, but these labels may be incorrect. Treating such noisy lab

safetyarxiv-cs-lg

Applications

One-Sided Verification Under Limited Expert Review

arXiv:2610.18089v1 Announce Type: new Abstract: AI has become increasingly helpful in scientific discovery, yet human expert review remains essential for those problems without reliable automated eval

applicationsarxiv-cs-ai

One-Step Prediction Fidelity Is a Poor Proxy for Planning Performance in JEPA World Models

arXiv:2610.19375v1 Announce Type: new Abstract: Latent world models are routinely selected, checkpointed and ablated on one-step prediction loss, on the assumption that a model which predicts the next

researcharxiv-cs-lg

Output Discipline Sets the Floor: A Falsification-Guided, Model-Agnostic Agent Protocol for Auditable Hypothesis-Driven Decision Making

arXiv:2610.18298v1 Announce Type: new Abstract: What binds large language models on auditable, hypothesis-driven decision tasks, where every candidate can be checked against the evidence? On the publi

Phantom Merge: When Your Large Language Model Agents Pick One but Tell You About Another

arXiv:2610.16141v1 Announce Type: new Abstract: LLM agents increasingly operate over multiple entities, where reliable response generation requires not only retrieving relevant evidence but also prese

PhysCodeGym: Benchmarking Coding Agents for Physically Verifiable 4D Reconstruction

arXiv:2610.17288v1 Announce Type: new Abstract: Building physical simulations from video requires coding agents to translate observations into executable models and refine them through tool use. We in

Physical Time as a Missing Dimension of Robot Policy Memory Dynamics

arXiv:2610.20524v1 Announce Type: new Abstract: Existing robot policy memories typically evolve with observation sequences, implicitly treating each update as the same amount of physical time. Yet ide

safetyarxiv-cs-ro

ProgAgent: Learning Tool Orchestration through Programmatic Training

arXiv:2610.14774v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly expected to orchestrate diverse tools over long-horizon tasks. Existing agents typically use Direct Tool C

Radioactive Feedback: Tracing Models Trained by Label-Only LLM Supervision

arXiv:2610.15685v1 Announce Type: new Abstract: Conventional distillation transfers capability through teacher-generated demonstrations. LLM-as-a-Judge knowledge distillation offers a distinct and inc

RadixFlow: Filling Execution Bubbles in Prefix-Cached Hybrid LLM Serving

arXiv:2610.18143v1 Announce Type: new Abstract: Prefix reuse improves LLM inference by sharing the KV cache of common prompt prefixes. This helps only if the scheduler can execute a matched branch in

RAN-JEPA: An Action-Conditioned World Model for Energy-efficient Mobile Network Control

arXiv:2610.17584v1 Announce Type: new Abstract: Operational mobile networks are rich in observations but poor in interventions: they continuously generate large volumes of passive telemetry, while con

RCCM: Diagnosing Gains and Harms in Reasoning-Prompt Transfer

model-releasesarxiv-cs-cl

arXiv:2610.19872v1 Announce Type: new Abstract: A reasoning prompt can improve one deployment and degrade another. RCCM (Reasoning-Control Capability Mismatch) examines these reversals by pairing answ

Receding-Horizon Next-Best-View Planning for Highly Occluded Target Perception

arXiv:2610.19082v1 Announce Type: new Abstract: Robotic systems that must acquire information about a specific target often face severe occlusion from sur- rounding structures or from the target itsel

researcharxiv-cs-ro

ReCreate: Reasoning and Creating Domain Agents Driven by Experience

arXiv:2601.11100v3 Announce Type: replace Abstract: Large Language Model (LLM) agents are reshaping the industrial landscape. However, most practical agents still rely on human-designed scaffolds, mak

Local Ai

Remember, Revisit, Revise: Scaffold Memory for Long-Horizon 3D Geometry

arXiv:2610.19674v1 Announce Type: new Abstract: Geometry foundation models estimate camera poses and 3D scene structure from video sequences, but maintaining consistent reconstructions over long horiz

local-aiarxiv-cs-cv

REPRO: Reflective Evolutionary Prompt Optimizer for Automated LLM Red-Teaming

arXiv:2610.15513v1 Announce Type: cross Abstract: Safety-aligned large language models (LLMs) remain vulnerable to adversarial prompts that bypass their safety training. Existing automated red-teaming

ReproBAIT: Can coding agents build their own biological AI tools?

arXiv:2610.19810v1 Announce Type: cross Abstract: Large language models are increasingly competent at leveraging scientific knowledge and existing computational tools to execute complex workflows. At

agentsarxiv-cs-lg

Riding the Fixed-Point Curve: Image Classification in the Continuous Thought Machine

arXiv:2610.19220v1 Announce Type: cross Abstract: The Continuous Thought Machine (CTM) is a recurrent network that attends to inputs and makes predictions based on the synchronization between its neur

researcharxiv-cs-cv

Robust Real-Time Monocular Human Pose Estimation with Insole-Based Sensor Fusion

arXiv:2610.19494v1 Announce Type: new Abstract: Real-time monocular human pose estimation is an important sensing component for human-centered intelligent systems. While recent camera-based methods ha

researcharxiv-cs-cv

RUBATO: Decoupled Clocks Allow Full-Duplex Models to Think as Fast as They Can

arXiv:2610.14299v1 Announce Type: cross Abstract: Recent Speech Full-Duplex (FDX) models are capable of engaging in naturalistic conversation by listening and speaking at the same time. Building on th

Safety for Whom? Benchmarking Understanding Beyond Refusal in Egocentric Multimodal Assistants

arXiv:2610.17728v1 Announce Type: new Abstract: Egocentric multimodal assistants see through the user's eyes and guide what the user does next, so their outputs directly shape physical actions, includ

Same Claim, Different Interpretations: How LLM Agents Resolve Ambiguity in Scientific Claims

model-releasesarxiv-cs-cl

arXiv:2610.20457v1 Announce Type: new Abstract: Co-scientists and agentic scientific claim verifiers are increasingly used to assess claims across the sciences. Natural-language scientific claims, wri

Scalable Robust MCTS with Anytime Deterministic Optimality Guarantees

arXiv:2610.15461v1 Announce Type: new Abstract: MDP planners often rely on nominal transition models that may differ from the true dynamics. Robust MDP planning accounts for this uncertainty through a

safetyarxiv-cs-ai

ScreenASR-Bench: A Tiered Benchmark for Speech Recognition with Visual Context

arXiv:2610.16260v1 Announce Type: new Abstract: Current ASR models perform well on general benchmarks but still struggle with acoustically ambiguous words. Visual context can aid recognition in screen

Local Ai

SearchMaster: Grounded and Regulated Self-Play for Search Agents

arXiv:2608.01822v2 Announce Type: replace Abstract: Training LLM-based search agents requires high-quality search data: tasks that demand genuine multi-hop retrieval and trajectories that use search t

local-aiarxiv-cs-ai

Selective Next-Location Prediction with MetaModelSelect: A Lightweight Post-Hoc Confidence Metamodel

arXiv:2610.19394v1 Announce Type: new Abstract: Next-location prediction can support proactive mobile-network decisions, but acting on unreliable predictions can waste resources. We apply MetaModelSel

researcharxiv-cs-lg

Self-Evolving Agent Harness

arXiv:2610.18813v1 Announce Type: new Abstract: Harness self-evolution revises an agent's software scaffolding, including prompts, memory, workflows, and executable components, while keeping model wei

safetyarxiv-cs-ai

Local Ai

Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models

arXiv:2610.19696v1 Announce Type: new Abstract: Multi-agent systems allow language models to solve tasks collaboratively by sharing complementary information. These systems typically communicate throu

local-aiarxiv-cs-cl

SkillScreen: High-Recall Screening of Malicious Agent Skills with Decision Models

arXiv:2610.17727v2 Announce Type: cross Abstract: Agent skills turn natural-language instructions into actions, creating a security challenge: legitimate automation and malicious behaviour can look de

Applications

SoK: Grading Open-Text Legal Answers, from Law School Exams to Trustworthy Rubric Judges

arXiv:2610.19992v1 Announce Type: new Abstract: Law students are examined through open-text answers: case solutions, essays and memos that apply the law to a set of facts. Such answers are expensive t

applicationsarxiv-cs-cl

SonoWorld: Towards Interactive Pose-Controllable Ultrasound World Models

arXiv:2610.19327v1 Announce Type: new Abstract: Robotic ultrasound can standardize image acquisition and reduce operator workload and variability, but policy learning remains limited by the lack of re

safetyarxiv-cs-cv

Sparse Autoencoders for Mechanistic Interpretability in NLP: A Critical Survey of Recoverability, Interpretability, and Utility

arXiv:2610.20139v1 Announce Type: new Abstract: Sparse autoencoders (SAEs) have emerged as a promising approach to mechanistic interpretability by decomposing language-model activations into sparse, p

researcharxiv-cs-cl

StartLux-Decision: AutoThink Decision Models with Reinforcement Learning for Calibrated Deliberative Decisions

arXiv:2610.17309v1 Announce Type: cross Abstract: We introduce StartLux-Decision, decision models from 0.8B to 27B parameters that combine efficient single-pass prediction with adaptive, confidence-gu

StateCraft: Augmenting Video World Models with Abstract-State Reasoning

model-releasesarxiv-cs-cv

arXiv:2610.19795v1 Announce Type: new Abstract: Video world models aim to simulate complex environments with realistic appearance and dynamics. Maintaining consistency over long interactions requires

StreamErase: Towards Ultra-Fast Streaming Video Erasing

model-releasesarxiv-cs-cv

arXiv:2610.20137v1 Announce Type: new Abstract: Current video object removal methods suffer from prohibitive latency (minutes per video) and non-causal processing that prevents real-time interaction.

Synthetic reaction data from transformation rules for graph-to-sequence pretraining in biodegradation pathway prediction

arXiv:2610.19640v1 Announce Type: new Abstract: The scarcity of observed biodegradation reactions constrains the training of product and pathway predictors. We present enviGraph, a framework that gene

researcharxiv-cs-lg

Applications

Taking the PULSE of Medical Reasoning: Step-Level Retrieval Timing for Knowledge-Grounded Inference

arXiv:2610.20292v1 Announce Type: new Abstract: Reliable medical reasoning requires LLMs to dynamically access external biomedical evidence to improve factual trustworthiness and reasoning accuracy. H

applicationsarxiv-cs-cl

Tutorials

Test-Time Scaling via Budgeted Multi-Attribute Verification

arXiv:2609.34322v2 Announce Type: replace Abstract: Verifying LLM-generated answers under a shared computational budget requires jointly deciding which candidates to inspect and which verification att

tutorialsarxiv-cs-ai

The Agent Knew: Internal States Forecast Unauthorized Actions and Stop Them, and the Lies That Follow

arXiv:2610.17301v1 Announce Type: new Abstract: Language-model agents sometimes act beyond their authorization and then misreport what they did. On CAPE-Bench, matched agent tasks in which one fact de

The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express

arXiv:2610.17158v1 Announce Type: new Abstract: When an LLM judge filters instruction-tuning data, its scoring format (binary verdict, 1--5 rating, 0--100 score, or comparison) is rarely justified or

safetyarxiv-cs-ai

Think in Strategies, Act with Tools: Traceable Agentic Optimization for Combinatorial Problems

arXiv:2610.14736v1 Announce Type: new Abstract: Optimization by PROmpting (OPRO) directly leverages Large Language Models (LLMs) to generate candidate solutions through natural-language prompting. How

To Reason but Not to Stop: Understanding and Recovering Pruned Large Reasoning Models

arXiv:2610.19934v1 Announce Type: new Abstract: Weight pruning can degrade the accuracy of large reasoning models (LRMs), yet it remains unclear whether it damages reasoning itself. We find that much

researcharxiv-cs-cl

Trace Evolution Bench: Measuring Whether Agents Improve from Their Own Environment Traces

arXiv:2610.15055v1 Announce Type: new Abstract: Self-evolving agents are usually scored by endpoint Gain: held-out success after an update minus success before. That difference can credit a frozen res