…Towards Standardized and Verifiable Evaluation of Multimodal Game Agents (2026) Vision2Web: A Hierarchical Benchmark for Visual Website Development with Agent Verification (2026) Towards Unified Multimodal Interleaved Generation via Group Relative Policy Optimization…
…through negative sample projection, maintaining diversity while outperforming existing methods on multiple benchmarks. AI-generated summary Reinforcement Learning with Verifiable Rewards (RLVR) enhances reasoning of Large Language Models (LLMs) but usually exhibits…
…on tokens where the teacher agrees with the verifier-correct direction and interpolates it where it disagrees. Experiments on competition-level mathematical reasoning benchmarks show that SG-OPD consistently outperforms standard OPD…
…Our two models, RoTS-7B and RoTS-32B, fine-tuned on our dataset, both demonstrate significant gains on GUI-RobustEval and traditional GUI benchmarks. Notably, RoTS-32B achieves state-of-the-art…
…A Document Parsing Benchmark for AI Agents (2026) Self-Ensembling Vision-Language Models for Chart Data Extraction (2026) Dr. DocBench: A Comprehensive Benchmark for Expert-Level and Difficult Document Parsing (2026) ABot…
…medium into a discrete puzzle? We identify this as the " verifiable reward trap ." While RLVR yields remarkable scores on standardized objective benchmarks, it systematically degrades the real-world conversational feel of audio…
…However, existing benchmarks are confined to single-stream paradigms, leaving a critical gap in evaluating online, cross-stream reasoning. To bridge this, we introduce X-Stream, the first benchmark dedicated to multi…
…math reasoning benchmarks, POISE matches DAPO while requiring less compute. Moreover, its value estimator shows similar performance to a separate LLM-scale value model and generalizes to various verifiable tasks. By leveraging…
…To study this question, we introduce UniKE, the first benchmark for cross-modality knowledge editing in UMMs, comprising 2,971 edit subjects spanning attribute and relation edits. Using VQA-based visual verification…
…https://arxiv.org/abs/2505.16483 Like OCC-RAG, CANOE improves contextual faithfulness without human annotation by synthesizing easily-verifiable short-form QA data across diverse tasks. The main difference is the…