qwen-2-5-vl-3b-230dc5ba·4 events·first seen Aliases: Qwen-2.5-VL-3B, Qwen2.5-VL-32B, Qwen2.5-32B
Researchers introduce PRISM, a single-layer logistic probe over LLM hidden states that distinguishes physically grounded danger (e.g., unsafe robot actions) from ordinary text-level content danger, showing these are separable signals in representations across multiple model families. PRISM achieves 86.2–87.7% accuracy on SafeAgentBench with substantially lower false-positive rates than same-scale LLM judges, which over-block safe tasks at 24.7–39.0% FPR. The authors also release PhysicalSafetyBench-1K (PSB-1K), a 1,000-pair contrastive benchmark for evaluating physical-risk detection without relying on explicit harm keywords. The work is relevant to safety in embodied AI and agentic systems where linguistic safety filters are insufficient.
RAGU is a new open-source modular GraphRAG engine that separates knowledge graph extraction from consolidation, using two-stage typed extraction, DBSCAN-backed deduplication, LLM summarization, and Leiden community detection. The authors train Meno-Lite-0.1, a 7B model specialized for language skills, which outperforms Qwen2.5-32B on knowledge-graph construction by 12.5% relative harmonic mean while matching it on English GraphRAG tasks. On GraphRAG-Bench (Medical), RAGU achieves evidence recall up to 0.84 versus ≤0.76 for competing systems. The system is MIT-licensed, pip-installable, and runs on a single GPU.
Researchers introduce PEEU (Planning Experience Exploration and Utilization), a training approach for small open-source multimodal LLMs that autonomously explores GUI environments to collect hindsight experience and synthesizes high-level training data for task planning. A 7B model trained with PEEU achieves 30.6% accuracy on real-world benchmarks, outperforming Qwen2.5-VL-32B. The paper also proposes TDHAF, a hierarchical analysis framework revealing that high-level task training yields stronger out-of-distribution generalization than mastering low-level atomic skills alone.
A new arXiv paper demonstrates that verifier-driven self-DPO, a common recipe for self-improving visual-language models, can silently degrade student model performance when the verifier's task-rubric accuracy is insufficient for the target task. Experiments on Qwen-3-VL-2B and Qwen-2.5-VL-3B across MathVista, MMMU, and BLINK show regressions of 3.4–10.9 percentage points below frozen baselines, with the counterintuitive finding that more accurate-but-still-wrong verifiers cause larger regressions than near-random ones. The authors provide a mechanistic explanation via a variance theorem for progress-gated replay and offer operational guidance: measure target-task rubric accuracy before running any verifier-driven loop and rank verifiers by task-specific quality rather than parameter count.