Can video generation models actually reason about the visual world — or do they just make things look plausible? These sound similar but are fundamentally different. Placing objects correctly, respecting physical laws, and solving multi-step tasks with coherent intermediate states: this is what "Visual Grounded Intelligence" means. VGI-Bench quantifies how far current models are from achieving it. Across 27 tasks and 810 instances, the strongest model — Seedance 2.0 — reaches only 51.0% overall. Three failure modes appear repeatedly: physical collapse (unrealistic deformations and object penetrations), rule violations (reaching plausible end states while ignoring constraints), and object inconsistency (losing temporal identity across frames). Open-source models lag far behind commercial ones, with the strongest open-source model at 19.1%. What the denoising analysis reveals is even more telling. Self-correction during intermediate generation steps occurs in fewer than 1% of transitions. Wrong-to-wrong transitions at mid-denoising reach 23.1%, and later denoising steps primarily refine early hypotheses rather than fixing errors. Fine-tuning on 1M synthetic samples improved planning and spatial skills but degraded temporal and physical capabilities — a tradeoff that points to deeper architectural constraints. VGI-Bench makes clear that visual reasoning remains a genuine unsolved frontier for video generation. Title: VGI-Bench: Probing Visual Intelligence in Video Generation Models URL: #VideoGeneration# #VisualReasoning#

TL;DR A new suite trains and evaluates "native visual reasoning," where visual generation itself is the medium of reasoning, using large-scale data and verifiable rewards. VLM-judge scores swung by up to 92.8% on the same video. Title: VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoning URL: Key points 📊 300 tasks, 1.25M training instances, ~3.47M images and 1.3M videos in a large-scale dataset 🎯 A deterministic scorer using classical CV (HSV color segmentation, OCR) hits 0.60+ human agreement, beating GPT-5.5 and Gemini-3.1-Pro 🌀 Coefficients-Preserving Sampling (CPS) keeps predicted and fresh noise coefficients balanced, stabilizing RL exploration 🚀 Transfer to V-ReasonBench jumps from 10.21 to 38.22 (+28.01 pts), the largest gain across seven external benchmarks 🔬 Counterfactual tests show removing input images drops scores by 90%, confirming visual trajectories matter more than text for reasoning ✅ Verifiable-reward RL beats VLM-reward RL by +7.9% in-domain What stands out is moving visual reasoning away from language dependence toward something trainable and verifiable in its own right. #VisualReasoning# #MultimodalAI#

Created with Seedance 2.5 ✨🎤 POV: you wake up barely functioning and somehow end the day under stage lights 😭✨✨ Prompt👇 CAMERA: Handheld DV 16mm camcorder aesthetic. CHASE records herself throughout every location, mostly holding the camera directly and occasionally setting it down for brief hands-free shots. Preserve natural hand tremors, crooked composition, slow autofocus correction, awkward zooms, accidental face cropping, and moments where the framing briefly loses her. The physical camcorder is never visible. LOOK: Authentic soft tape-image texture with mild blur, subtle electronic grain, glowing highlights, small automatic-exposure fluctuations, restrained contrast, and natural skin tones. The lighting changes realistically between locations: warm morning light in the apartment → cooler daylight inside the car → intense colorful lighting backstage and onstage. STYLE: A fast, energetic day-in-the-life montage with sharp cuts between locations and slightly accelerated transitional movement. Reflective voiceover plays over the footage instead of synchronized dialogue. The pace gradually builds from a quiet sleepy morning into a high-energy performance finale. Character CHASE — an exceptionally beautiful Instagram influencer and rising pop performer in her 20s. Long glossy dark-brown hair, striking symmetrical features, large expressive eyes, flawless glowing skin, and soft pink lips. Slim, toned figure. Her outfit changes with each location: cozy loungewear at home, stylish casual clothing during the ride, and a glamorous performance outfit for the final scene. Each outfit fully covers her arms and torso. Setting Progression Stylish apartment bedroom in the morning → luxury van interior during the day → busy backstage area and concert stage at night. Storyboard — 15 seconds, 5 cuts (~3s, bedroom, camera resting on a table, soft morning light) She slowly sits up, stretches, fixes her messy hair, and hurriedly packs a small bag. VOICEOVER (CHASE): “Some mornings begin before I’m even fully awake.” (~3s, van interior, handheld selfie footage, natural window light) She sits beside the window as sunlight moves across her face, casually checking messages on her phone. VOICEOVER (CHASE): “I spend so much time on the road, this place almost feels like home.” (~3s, van interior, macro detail shot) Close-up of her fingers choosing a song from a playlist while passing sunlight flickers across the screen and her hand. No voiceover, only natural road noise. (~3s, backstage, fast handheld montage) Quick glimpses of makeup brushes, hair styling, costume adjustments, staff rushing past, and her taking one focused breath before going onstage. VOICEOVER (CHASE): “And then everything suddenly starts moving at once.” (~3s, stage, wide shot moving into a close-up, energetic finale) Bright stage lights ignite as she steps through the curtains. The camera catches her silhouette, then a brief confident smile before the image cuts to black. VOICEOVER (CHASE): “This moment is why I keep doing it.” Try it on Ima Studio 👉 #Seedance25# #AIVideo# #ImaStudio# #VisualStorytelling#