All papers

Introduces a benchmark combining natural, scripted, and synthetic audio to evaluate phonetic error detection in speech models.

The authors introduce a benchmark to audit mathematical claims and repair errors using an automated witness verification checker.

This paper introduces an evaluation index to measure whether machine translation of emergency communication preserves critical safety actionability.

The paper proposes a training-free method combining global and regional representations across encoder layers for few-shot vision-language adaptation.

The paper introduces the Korean Benchmark Suite to evaluate foundation models across six natural language capabilities.

This paper presents a label differential privacy framework to protect survey-derived ordinal preference data during language model alignment.

Introduces an AND/OR graph proof search algorithm optimizing customizable objectives like length and compute cost during automated formal theorem proving.

The paper presents a contrastive learning framework that handles false-negative pairs during zero-shot chest X-ray vision-language pre-training.

The authors introduce a benchmark to audit mathematical claims and repair errors using an automated witness verification checker.

Models repeated games with endogenous delegation to analyze trade-offs between economies of scale and institutional accountability.

The paper examines conceptual parallels between measurement science and machine learning uncertainty evaluation to propose a more unified assessment framework.

The paper introduces an image-conditioned evidence refinement method to improve fine-grained compositional matching in vision-language models.

This paper compares observational feature discovery scores in sparse autoencoders against actual causal intervention effects on model outputs.

The paper introduces Swiss4Arg, a multilingual argument mining dataset based on official Swiss voting brochures across four national languages.

Presents an end-to-end joint-embedding predictive architecture that combines world modeling and action flow matching for robot planning and control.

Introduces a robotic system that converts language instructions and visual inputs into physical light projection actions in spatial environments.

The paper presents a recursive self-improvement framework using scientific agent swarms that mutate hypotheses and recombine code based on experimental feedback.

This paper introduces Lüshi-1840, a pre-modern Chinese language model trained on historical texts to evaluate knowledge cutoffs and hindsight bias.

The paper evaluates how commercial language models value human time by measuring recommended delay compensation amounts across 193 different countries.

Improves on-policy distillation for diffusion models by restricting feature alignment supervision to deep layers and early denoising steps.

The paper provides an analytical framework for how multimodal agents form, represent, and adapt memory over long-horizon tasks.

The paper introduces a bidirectional gloss-pair prediction task to evaluate semantic relation knowledge across multiple languages in language models.

Demonstrates that discrete graph diffusion models exhibit conditional memorization when reverse trajectories are conditioned on corrupted training graphs.

Presents an agentic multimodal framework using a plan-act-observe loop to generate structured, protocol-compliant chest X-ray radiology reports.

JevGPT combines fast local evidence assessments with deliberative planning to improve autonomous research agents operating under fixed retrieval budgets.

Employs a bilevel Stackelberg formulation to jointly optimize classification accuracy and confidence calibration without separate calibration datasets.

Introduces a Ukrainian meme dataset and evaluates vision-language models on detecting persuasive and propaganda techniques.

Proposes an AI-generated video quality assessment method using visually grounded captions instead of relying solely on source text prompts.

Studies how fine-tuning helps tool-using agents learn to withdraw and repair actions when environment assumptions change.

The paper proposes an iterative self-improvement method combining trajectory rewriting and policy refinement for terminal-based coding agents.

The paper introduces a benchmark framework evaluating LLM-generated quantum software unit tests against independent ground-truth checkers.

Proposes a training-free closed-form method to refine target object retrieval in 3D Gaussian language maps while preserving open-vocabulary queries.

The authors propose a post-training method enabling autonomous language model agents to organize experiments and carry out machine learning engineering tasks.

The paper presents a multilingual benchmark of minimal pairs to evaluate language model sensitivity to negative polarity items across 33 languages.

Introduces MAGMA-Bench to evaluate long-horizon embodied agent reasoning during dynamic task execution failures and multi-robot coordination.

Distills multi-view visual cues into a latent flow matching network to improve point cloud upsampling at inference time.

Presents a multi-agent framework that decouples research planning, subquestion investigation, and factual verification via a shared semantic workspace.

The authors introduce a benchmark evaluating egocentric video understanding structured around an interactive loop of user perception, action, and world feedback.

Introduces an execution-grounded dataset and method to repair geometric and engineering constraint violations in LLM-generated industrial scene programs.

Combines visual tracking and vocal fingerprinting to identify and monitor individual dogs across real-world multi-dog video recordings.

The paper presents a vision-language policy combining pre-aligned representations and multi-stage learning to improve robotic person tracking among distractors.

AnnalsChallenge provides 50 formalized mathematics theorem statements in Lean 4 drawn from papers published in the Annals of Mathematics.

Audits audio-language model hallucination mitigations by testing localized acoustic evidence ablation against baseline methods across model architectures.

Provides region-specific reward modeling to fix incorrect wording and glyph errors in visual text generation models.

The paper evaluates seven language models on estimating soil organic carbon concentrations, calculating carbon stock, and comparing agricultural paddock pairs.

Proposes a benchmark and probing methodology based on attribution theory to examine how language models assign responsibility and blame.

Presents a textbook providing an overview of the concepts, architectures, and interactions of agentic large language models.

The authors introduce CharXiv, a benchmark of realistic scientific charts designed to test descriptive and reasoning capabilities in multimodal language models.

The paper introduces a dataset and benchmark to evaluate vision-language models on recognizing and answering questions about Philippine cultural heritage sites.

Proposes an adaptive multi-source steering decoding method to mitigate hallucinations in vision-language models by balancing visual and textual context contributions.

Introduces a multimodal embodied agent policy learning framework using trajectory construction, distillation, and token-level evidence routing.

The authors introduce a training loop where multimodal models use text and visual abilities alongside program execution to generate their own training data.

The paper introduces an autonomous multi-agent framework that translates natural language prompts into playable browser games with automated verification.

The paper evaluates the impact of sarcasm on Turkish hate speech detection models using a targeted external test dataset.

Introduces a training-free inference method using Gaussian optimal transport to reduce object hallucination in vision-language models without suppressing valid representations.

The paper introduces a benchmark and evaluation framework assessing context-sensitive code-switching by language models in creative writing tasks.

This paper analyzes visual token utilization in high-resolution vision-language models to optimize token pruning across network layers.

Analyzes interaction failures between KL regularization and rewards in group policy optimization and proposes a calibrated surrogate objective.

Introduces HOA, a multi-session LLM agent framework coordinating review, wiki knowledge, and verification for complex Olympiad problem solving.

The paper introduces prompt optimization and recovery measurement tools to evaluate the limits of transmitting concepts between scaling language models.