Many users praised the Flash-BoN paper for its diversity plots and extensive timing experiments in timed image generation scaling, while thanking contributors for the explainer video.
Based on 3 visible X reactions from 2 accounts; directional sample.
Ask a question below.
Published answers will appear here.
This paper involved an unreasonable number of timing runs, verifier arguments, and plots that changed depending on the x-axis. Very lucky to have gone down this rabbit hole with @reza_shkv, @RisingSayak, @ywen99, Heng, Yizheng, @tomgoldsteincs, and @gowthami_s. Paper + code: https://flash-bon.github.io/
Also, a big thanks to @RisingSayak for putting together the lovely explainer video 🧑🎓
Most inference time scaling approaches are either naive or use fixed recipe for all the prompts. Our #ECCV work combines 3 knobs to get a good generation out fast!! Paper has lots of analysis on issues with current approaches, how to do inference scaling right and also how to pick a good rater for the generations. I had lot of fun working on this with @RawalRuchit and other awesome collaborators.
We tested this on few open source models, obv this should scale for others too as long as architectures don’t diverge too much! Paper: https://arxiv.org/abs/2607.04461 Code: https://github.com/flash-bon/flash-bon
This paper involved an unreasonable number of timing runs, verifier arguments, and plots that changed depending on the x-axis. Very lucky to have gone down this rabbit hole with @reza_shkv, @RisingSayak, @ywen99, Heng, Yizheng, @tomgoldsteincs, and @gowthami_s. Paper + code: https://flash-bon.github.io/
Most inference time scaling approaches are either naive or use fixed recipe for all the prompts. Our #ECCV work combines 3 knobs to get a good generation out fast!! Paper has lots of analysis on issues with current approaches, how to do inference scaling right and also how to pick a good rater for the generations. I had lot of fun working on this with @RawalRuchit and other awesome collaborators.
We tested this on few open source models, obv this should scale for others too as long as architectures don’t diverge too much! Paper: https://arxiv.org/abs/2607.04461 Code: https://github.com/flash-bon/flash-bon
Many users praised the Flash-BoN paper for its diversity plots and extensive timing experiments in timed image generation scaling, while thanking contributors for the explainer video.
Based on 3 visible X reactions from 2 accounts; directional sample.
Ask a question below.
Published answers will appear here.