Twice Upon a Time: Complementary Representations for Time-Series Reasoning
A Qwen3-VL-8B dual-view time-series model lifts TSRBench accuracy from 44.5% to 53.2%.
Stories per day · 30 days
3 RSS
18 stories, newest first · one card per story
Stories tagged #qwen3-vl
AtomicChat/Qwen-Image-2.1-Turbo-Uncensored-GGUF — new model trending #30 on Hugging Face
AtomicChat published abliterated GGUF weights of the Qwen-Image 2.1 Turbo text encoder, sharply cutting chat refusals.
23
FastBench: Can Streaming VLMs Perceive High-Dynamic Real-World Streams?
FastBench shows streaming VLMs miss fast events; Gemini-3.5-Flash scores only 50.7%.
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models
SpaceCast-Bench tests predictive spatial reasoning; the best of 21 models scores 58.0% versus 87.2% for humans.
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models
SpaceCast-Bench finds VLMs weak at predictive spatial reasoning; the best model scores 58 percent versus 87.2 percent for humans.
OneSearch-VL: Unified Multimodal Deep Research Agent for Image and Video
OneSearch-VL-8B beats tool-using Qwen3-VL-8B by up to 20.2 points on new visual research benchmarks.
Foresight: planning future perception in streaming VLMs without retraining
FORESIGHT lets frozen Qwen3-VL-8B plan future perception in streaming video without retraining.
AI Search is now generally available
Cloudflare AI Search is generally available with native image embeddings, PDF OCR, and billing from November 1, 2026.
6
GUI-HARVEST: Self-Improving GUI Agents through Evidence-Driven Harness Evolution
GUI-HARVEST evolves GUI agent harnesses, lifting frozen Qwen3-VL by 12.33 points on OSWorld.
CANOPY: Adaptive-Granularity Evidence Compression for Multimodal RAG
CANOPY compresses multimodal RAG evidence at mixed granularities, cutting reader tokens without hurting answer accuracy.
Ranking-Aware Prompt Optimization for Multimodal Clinical Diagnosis
Ranking-aware prompt evolution lifts clinical AUROC on Qwen3-VL-8B and MedGemma-4B versus accuracy-based search.
EgoTools: Towards Tool-Centric Reasoning in Real-World Egocentric Videos
EgoTools adds 100 hours of egocentric tool-use video and a 1,000-question benchmark where current VLMs still struggle.
MAD-Guard: Controlled Study of Autoregressive Generation versus Direct Decision Interfaces for Closed Multimodal Forensic Tasks
MAD-Guard shows direct decision heads outperform autoregressive decoding for closed multimodal forgery forensics.
Program-Verified Self-Evolution for Vision-Language Models
VQS generates program-verified vision questions, labeling 94% correctly and lifting Qwen3-VL by up to 3.84 points.
Cross-Modal Attention Acts as a Frequency Filter: Why Verbose Prompts Improve Robustness in Vision-Language Models
Study explains why verbose prompts make vision-language models more robust to image corruption, framing cross-modal attention as a frequency filter.
Can Edge-Deployable Vision-Language Models Identify Species?
Evaluation of 2-8B VLMs (Qwen3-VL, Gemma3) against BioCLIP on camera-trap species ID shows all models degrade sharply on field imagery.
MindTopo: Can Foundation Models Reason in Topological Space?
MindTopo benchmark with 11,030 topological tasks shows 14 multimodal LLMs reason better than they plan and remain far below human performance.
EmbodiedSkills: A Unified Framework for Orchestrating, Training, and Deploying VLA Agents
Researchers introduce EmbodiedSkills, a framework treating VLA skill decisions as verified execution proposals, reaching 86.2% success on RoboTwin 2.0.