Üretim4
görselmodel·hf-papers·2 gün önce·30 oy·+1 kaynak
Unified multimodal models can both look at and render images, so in principle they can repair their own generations: diagnose what an image gets wrong, revise it, observe the result, and diagnose again. Whether a revision helps is known only after it is rendered, so the reflection text and the image generation must be learned jointly, over the whole loop. Supervised fine-tuning (SFT) on reflection trajectories gives a cold start but does not find the high-success repair paths, and naive RL that optimizes only the renderer or only one head leaves most of the gain untapped. We introduce…
görselmodel·hf-papers·2 gün önce·50 oy
Most modern multimodal large language models (MLLMs) build on a pretrained visual encoder that provides a strong visual prior. Encoder-free MLLMs instead learn visual representations directly from raw pixels, offering a simple and unified architecture, but their scaling behavior has not been systematically characterized. To fill this gap, we compare scaling laws for encoder-free and encoder-based MLLMs and report three main findings: (1) Removing the visual encoder shifts the compute-optimal allocation for the multimodal objective toward larger models, while leaving that for text nearly…
görsel3D·reddit-stablediffusion·7 saat önce
... if you want to use the hybrid model. Ref2VA has certain quality issues and struggles with mid‑distance subjects and colors, it's a known issue the Minimax team has acknowledged. Hybrid gives you FL2VA's quality with Ref2VA's flexibility. But some capabilities take a hit, video referencing is one. The character swap LoRA brings it back and makes it even better. Even Ref2VA strongly benefits from it according to my tests. At 88% denoise plus LoRA, the accuracy is even better but it may not be suitable on all scenarios. It depends on the video. I'm including more tests in the comments…
görsel3D·reddit-stablediffusion·3 saat önce
Create a reference image of the face (showing various angles) arranged in a 2x2 grid (in ComfyUI, this is easily done using the "Load Images & Compose (obvpm)" node: https://github.com/chanon/comfyui-obvpm ). In the prompt (created by AI), replace the square brackets with your own prompt: Task: Analyze the provided reference image containing 4 panels. These are four different angles, expressions, and lighting conditions of the same character. Your goal is to merge their facial features into one single, consistent, canonical face. Step 1 (Identity Extraction): Extract the core identity markers…
Modeller & Araştırma4
↑ yükselişte31 oy/saataraçmodelşirket·hackernews·17 saat önce·541 oy
model·openai·18 saat önce
Our early guidelines for safety cases in frontier AI training cover technical safeguards, operational practices, and investigating misalignment incidents
↑ yükselişteHN'de GPT-6 14→25 hikâyemodelşirket·openai·2 gün önce
GPT-6 Astra completed a 50-tab tax workbook twice as fast as GPT-5.6 Sol, and its stronger understanding of user intent gives Basis more confidence in real-world use.
model·simonwillison·15 saat önce
Claude Sonnet 5.5 New Sonnet model from Anthropic today. They say it "runs 30%+ faster, and costs up to 30% less for most work" - it's priced the same as Sonnet 5 but appears to beat it on every benchmark, and should be cheaper to run as well. Here are some pelicans riding bicycles . Sonnet 5.5 suffered from the same bug as Opus 5.5 : the "max" thinking effort pelican thought for 128,000 tokens (at a cost of $1.28) before running out of tokens and failing to produce an SVG. Here's the pelican it gave me for thinking effort "xhigh", at a cost of 5.74 cents and taking 41 seconds: Sonnet 5.5…