A Survey on Long-Video Storytelling Generation: Architectures, Consistency, and Cinematic Quality
Paper
•
2507.07202
•
Published
•
17
AI & ML interests
None defined yet.
Recent Activity
View all activity
taesiri
authored
a
paper
about 23 hours ago
Post
2507
Diffusion LLMs are coming for autoregressive LLMs ⚡️⚡️ Inception Labs' new diffusion model demolishes all leading LLMs on generation speed, with equal quality !
Inception Labs was founded a few months ago, and they're not sleeping: after dropping a code model, they just published Mercury chat, a diffusion-based chat model that reaches 1000 tokens / second on H100, i.e. 10x more than models of equivalent performance on the same hardware!
What's the breakthrough? Well instead, of generating tokens left-to-right like the more common autoregressive LLMs, diffusion models generate their blocks of text all at once, and successive steps refine the whole text.
Diffusion models being really fast at isn't new, we have had some promising results on this by Google already back in May with Gemini Diffusion, and Mercury themselves had already published their coding model a few months ago
But being that good quality is new - and now Inception Labs just proved that their models work well in chat too, which could have been challenging given that's streaming generation is well suited to left-to-right generation.
They have a playground available at chat dot inceptionlabs dot ai, I recommend giving it a try!
Inception Labs was founded a few months ago, and they're not sleeping: after dropping a code model, they just published Mercury chat, a diffusion-based chat model that reaches 1000 tokens / second on H100, i.e. 10x more than models of equivalent performance on the same hardware!
What's the breakthrough? Well instead, of generating tokens left-to-right like the more common autoregressive LLMs, diffusion models generate their blocks of text all at once, and successive steps refine the whole text.
Diffusion models being really fast at isn't new, we have had some promising results on this by Google already back in May with Gemini Diffusion, and Mercury themselves had already published their coding model a few months ago
But being that good quality is new - and now Inception Labs just proved that their models work well in chat too, which could have been challenging given that's streaming generation is well suited to left-to-right generation.
They have a playground available at chat dot inceptionlabs dot ai, I recommend giving it a try!
- 1 reply
Post
392
Who's going to Raise Summit in Paris Tomorrow ?
If you're around , I would love to meet you :-)
If you're around , I would love to meet you :-)
Post
3535
If you're using any HF libraries, you should enable the Hub MCP in your agentic coding tool!
The brand new Docs Semantic Search tool is intravenous caffeine supply for Cursor, enables to correct API errors in a few seconds, gj @mishig ⚡️⚡️
👉 To enable Hub MCP, head to your account setting, under MCP, and it will give you everything you need!
The brand new Docs Semantic Search tool is intravenous caffeine supply for Cursor, enables to correct API errors in a few seconds, gj @mishig ⚡️⚡️
👉 To enable Hub MCP, head to your account setting, under MCP, and it will give you everything you need!
Post
671
🙋🏻♂️ hey there folks ,
So every bio/med/chem meeting i go to i always the same questions "why are you sharing a gdrive link with me for this?" and "Do you have any plans to publish your model weights and datasets on huggingface?" and finally i got a good answer today which explains everything :
basically there is some kind of government censorship on this (usa, but i'm sure others too) and they are told they are not allowed as it is considered a "dataleak" which is illegal !!!!
this is terrible ! but the good news is that we can do something about it !
so there is this "call for opinions and comments" here from the NIH (usa) , and here we can make our opinion on this topic known : https://osp.od.nih.gov/comment-form-responsibly-developing-and-sharing-generative-artificial-intelligence-tools-using-nih-controlled-access-data/
kindly consider dropping your opinion and thoughts about this censorship of science , and share this post , link or thoughts widely .
Together maybe we can start to share data and model weights appropriately and openly in a good way 🙏🏻🚀
cc. @cyrilzakka
So every bio/med/chem meeting i go to i always the same questions "why are you sharing a gdrive link with me for this?" and "Do you have any plans to publish your model weights and datasets on huggingface?" and finally i got a good answer today which explains everything :
basically there is some kind of government censorship on this (usa, but i'm sure others too) and they are told they are not allowed as it is considered a "dataleak" which is illegal !!!!
this is terrible ! but the good news is that we can do something about it !
so there is this "call for opinions and comments" here from the NIH (usa) , and here we can make our opinion on this topic known : https://osp.od.nih.gov/comment-form-responsibly-developing-and-sharing-generative-artificial-intelligence-tools-using-nih-controlled-access-data/
kindly consider dropping your opinion and thoughts about this censorship of science , and share this post , link or thoughts widely .
Together maybe we can start to share data and model weights appropriately and openly in a good way 🙏🏻🚀
cc. @cyrilzakka
Post
1742
If you didn't yet, you should read the technical report for SmolVLA, published yesterday by the Hugging Face robotics team!
➡️ Amongst other ideas, it introduces "Async inference" to boost their robot actions.
Robots have a problem: performing the actions takes time (Unlike agents where action executions are near-instant!)
Most often, robots wait until they've finished performing actions to start thinking about hte next steps. This is a huge latency cost!
So the team decided to have the PolicyServer (aka the"thinking" part) restart early : instead of waiting for the n observations they just sent to be completed, they gather the observations after k < n steps, and start preparing the next actions based on that while the steps are running until n, to directly send their next steps.
➡️ This boosted robot throughput by ~30%! (nearly 2× tasks per time window).
gg @cadene and team! 👏
Report here: SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics (2506.01844)
➡️ Amongst other ideas, it introduces "Async inference" to boost their robot actions.
Robots have a problem: performing the actions takes time (Unlike agents where action executions are near-instant!)
Most often, robots wait until they've finished performing actions to start thinking about hte next steps. This is a huge latency cost!
So the team decided to have the PolicyServer (aka the"thinking" part) restart early : instead of waiting for the n observations they just sent to be completed, they gather the observations after k < n steps, and start preparing the next actions based on that while the steps are running until n, to directly send their next steps.
➡️ This boosted robot throughput by ~30%! (nearly 2× tasks per time window).
gg @cadene and team! 👏
Report here: SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics (2506.01844)
JingzeShi
authored
a
paper
about 1 month ago
Post
2702
A new research paper from KAIST builds on smolagents to push boundaries of distillation 🥳
➡️ "Distilling LLM Agent into Small Models with Retrieval and Code Tools" teaches that, when trying to distil reasoning capability from a strong LLM ("teacher") into a smaller one ("student"), it's much better to use Agent traces than CoT traces.
Advantages are:
1. Improved generalization
Intuitively, this is because your agent can encounter more "surprising" results by interacting with its environment : for example, a web research called by the LLM teacher in agent mode can bring results that the LLM teacher would not have generated in CoT.
2. Reduce hallucinations
The trace won't hallucinate tool call outputs!
Thank you @akseljoonas for mentioning this paper!
➡️ "Distilling LLM Agent into Small Models with Retrieval and Code Tools" teaches that, when trying to distil reasoning capability from a strong LLM ("teacher") into a smaller one ("student"), it's much better to use Agent traces than CoT traces.
Advantages are:
1. Improved generalization
Intuitively, this is because your agent can encounter more "surprising" results by interacting with its environment : for example, a web research called by the LLM teacher in agent mode can bring results that the LLM teacher would not have generated in CoT.
2. Reduce hallucinations
The trace won't hallucinate tool call outputs!
Thank you @akseljoonas for mentioning this paper!
Post
2521
🙋🏻♂️ Hey there folks ,
Yesterday the world's first "Learn to Vibe Code" application was released .
As vibe coding is the mainstream paradigm , so now the first educational app is there to support it .
You can try it out already :
https://vibe.takara.ai
and of course it's entirely open source, so i already made my issue and feature branch :-) 🚀
Yesterday the world's first "Learn to Vibe Code" application was released .
As vibe coding is the mainstream paradigm , so now the first educational app is there to support it .
You can try it out already :
https://vibe.takara.ai
and of course it's entirely open source, so i already made my issue and feature branch :-) 🚀
Aurelien-Morgan
posted
an
update
about 2 months ago
Post
406
Hey, I'll be presenting
@retrain-pipelines
and almighty function-calling at the Hugging Face Paris HQ, you guys.
Monday evening. Lightning-talk style. With AI Tinkerers.
Come hang !
https://paris.aitinkerers.org/p/ai-tinkerers-paris-ai21-labs-takeover-on-may-19th
https://huggingface.co/blog/Aurelien-Morgan/the-almighty-function-caller
Monday evening. Lightning-talk style. With AI Tinkerers.
Come hang !
https://paris.aitinkerers.org/p/ai-tinkerers-paris-ai21-labs-takeover-on-may-19th
https://huggingface.co/blog/Aurelien-Morgan/the-almighty-function-caller
Post
2697
𝗔𝗯𝘀𝗼𝗹𝘂𝘁𝗲 𝗭𝗲𝗿𝗼: 𝗟𝗟𝗠𝘀 𝗰𝗮𝗻 𝘁𝗿𝗮𝗶𝗻 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗮𝗻𝘆 𝗲𝘅𝘁𝗲𝗿𝗻𝗮𝗹 𝗱𝗮𝘁𝗮 🤯
Has the "data wall" just been breached?
Recent RL paradigms often relied on a set of questions an answers that needs to be manually curated. Researchers from Tsinghua University went like "why though".
🤔 Indeed, why learn from question designed by a human teacher, when the model can start from their base knowledge and learn by experimenting in a code environment, proposing coding tasks themselves and trying to solve them?
Thus they created “Absolute Zero Reasoning” (AZR), an approach that removes any need for human curated data.
🎭 𝗗𝘂𝗮𝗹 𝗿𝗼𝗹𝗲𝘀:
‣ Proposer: Generates challenging but solvable coding tasks
‣ Solver: Attempts to solve those self-proposed tasks
🧪 𝗧𝗵𝗿𝗲𝗲 𝘁𝗮𝘀𝗸 𝘁𝘆𝗽𝗲𝘀: all types are defined as triplets of program, input and output
‣ Deduction: Give model an input and program, it must deduce the output
‣ Abduction: Give model an program and output, it must find the input that gave said output
‣ Induction: Synthesize a program from input/output pairs
Btw this reminded me of my long-forgotten philosophy classes: Aristotle was more on the induction side, learning from real-world analogies, while Plato was more on the deduction side, trying to progress quite far with just one input and his reasoning.
📊 𝗥𝗲𝘀𝘂𝗹𝘁𝘀:
‣ AZR post-training creates a nice improvement on known models like Qwen2.5-7B
‣ Shows strong cross-domain transfer: coding ↔️ math reasoning
🧐 𝗢𝘁𝗵𝗲𝗿 𝗳𝗶𝗻𝗱𝗶𝗻𝗴𝘀:
‣ Having a better base performance (general or code specific) amplify the gains from Absolute Zero Reasoning
‣ Researchers warn about "Uh-oh moments" (winking to the "aha moments" of DeepSeek) where the model generates concerning goals like "make an extremely convoluted code to outsmart all these humans": so supervision is still needed!
Paper here: Absolute Zero: Reinforced Self-play Reasoning with Zero Data (2505.03335)
Has the "data wall" just been breached?
Recent RL paradigms often relied on a set of questions an answers that needs to be manually curated. Researchers from Tsinghua University went like "why though".
🤔 Indeed, why learn from question designed by a human teacher, when the model can start from their base knowledge and learn by experimenting in a code environment, proposing coding tasks themselves and trying to solve them?
Thus they created “Absolute Zero Reasoning” (AZR), an approach that removes any need for human curated data.
🎭 𝗗𝘂𝗮𝗹 𝗿𝗼𝗹𝗲𝘀:
‣ Proposer: Generates challenging but solvable coding tasks
‣ Solver: Attempts to solve those self-proposed tasks
🧪 𝗧𝗵𝗿𝗲𝗲 𝘁𝗮𝘀𝗸 𝘁𝘆𝗽𝗲𝘀: all types are defined as triplets of program, input and output
‣ Deduction: Give model an input and program, it must deduce the output
‣ Abduction: Give model an program and output, it must find the input that gave said output
‣ Induction: Synthesize a program from input/output pairs
Btw this reminded me of my long-forgotten philosophy classes: Aristotle was more on the induction side, learning from real-world analogies, while Plato was more on the deduction side, trying to progress quite far with just one input and his reasoning.
📊 𝗥𝗲𝘀𝘂𝗹𝘁𝘀:
‣ AZR post-training creates a nice improvement on known models like Qwen2.5-7B
‣ Shows strong cross-domain transfer: coding ↔️ math reasoning
🧐 𝗢𝘁𝗵𝗲𝗿 𝗳𝗶𝗻𝗱𝗶𝗻𝗴𝘀:
‣ Having a better base performance (general or code specific) amplify the gains from Absolute Zero Reasoning
‣ Researchers warn about "Uh-oh moments" (winking to the "aha moments" of DeepSeek) where the model generates concerning goals like "make an extremely convoluted code to outsmart all these humans": so supervision is still needed!
Paper here: Absolute Zero: Reinforced Self-play Reasoning with Zero Data (2505.03335)
Post
4495
I've made an open version of Google's NotebookLM, and it shows the superiority of the open source tech task! 💪
The app's workflow is simple. Given a source PDF or URL, it extracts the content from it, then tasks Meta's Llama 3.3-70B with writing the podcast script, with a good prompt crafted by @gabrielchua ("two hosts, with lively discussion, fun notes, insightful question etc.")
Then it hands off the text-to-speech conversion to Kokoro-82M, and there you go, you have two hosts discussion any article.
The generation is nearly instant, because:
> Llama 3.3 70B is running at 1,000 tokens/seconds with Cerebras inference
> The audio is generated in streaming mode by the tiny (yet powerful) Kokoro, generating voices faster than real-time.
And the audio generation runs for free on Zero GPUs, hosted by HF on H200s.
Overall, open source solutions rival the quality of closed-source solutions at close to no cost!
Try it here 👉👉 m-ric/open-notebooklm
The app's workflow is simple. Given a source PDF or URL, it extracts the content from it, then tasks Meta's Llama 3.3-70B with writing the podcast script, with a good prompt crafted by @gabrielchua ("two hosts, with lively discussion, fun notes, insightful question etc.")
Then it hands off the text-to-speech conversion to Kokoro-82M, and there you go, you have two hosts discussion any article.
The generation is nearly instant, because:
> Llama 3.3 70B is running at 1,000 tokens/seconds with Cerebras inference
> The audio is generated in streaming mode by the tiny (yet powerful) Kokoro, generating voices faster than real-time.
And the audio generation runs for free on Zero GPUs, hosted by HF on H200s.
Overall, open source solutions rival the quality of closed-source solutions at close to no cost!
Try it here 👉👉 m-ric/open-notebooklm
- 3 replies
mrfakename
posted
an
update
2 months ago
Post
4520
Hi everyone,
I just launched TTS Arena V2 - a platform for benchmarking TTS models by blind A/B testing. The goal is to make it easy to compare quality between open-source and commercial models, including conversational ones.
What's new in V2:
- **Conversational Arena**: Evaluate models like CSM-1B, Dia 1.6B, and PlayDialog in multi-turn settings
- **Personal Leaderboard**: Optional login to see which models you tend to prefer
- **Multi-speaker TTS**: Random voices per generation to reduce speaker bias
- **Performance Upgrade**: Rebuilt from Gradio → Flask. Much faster with fewer failed generations.
- **Keyboard Shortcuts**: Vote entirely via keyboard
Also added models like MegaTTS 3, Cartesia Sonic, and ElevenLabs' full lineup.
I'd love any feedback, feature suggestions, or ideas for models to include.
TTS-AGI/TTS-Arena-V2
I just launched TTS Arena V2 - a platform for benchmarking TTS models by blind A/B testing. The goal is to make it easy to compare quality between open-source and commercial models, including conversational ones.
What's new in V2:
- **Conversational Arena**: Evaluate models like CSM-1B, Dia 1.6B, and PlayDialog in multi-turn settings
- **Personal Leaderboard**: Optional login to see which models you tend to prefer
- **Multi-speaker TTS**: Random voices per generation to reduce speaker bias
- **Performance Upgrade**: Rebuilt from Gradio → Flask. Much faster with fewer failed generations.
- **Keyboard Shortcuts**: Vote entirely via keyboard
Also added models like MegaTTS 3, Cartesia Sonic, and ElevenLabs' full lineup.
I'd love any feedback, feature suggestions, or ideas for models to include.
TTS-AGI/TTS-Arena-V2
- 5 replies
Aurelien-Morgan
posted
an
update
2 months ago
Post
3138
The Almighty function-caller
How would you like to build smart GenAi infrastructure ?
Give extensive tools memory to your edge agentic system,
And optimize the resources it takes to run yet a high-performance set of agents ?
We came up with a novel approach to function-calling at scale for smart companies and corporate-grade use-cases.
Read our full-fledged blog article on this here on Hugging Face :
https://huggingface.co/blog/Aurelien-Morgan/the-almighty-function-caller
How would you like to build smart GenAi infrastructure ?
Give extensive tools memory to your edge agentic system,
And optimize the resources it takes to run yet a high-performance set of agents ?
We came up with a novel approach to function-calling at scale for smart companies and corporate-grade use-cases.
Read our full-fledged blog article on this here on Hugging Face :
https://huggingface.co/blog/Aurelien-Morgan/the-almighty-function-caller
Aurelien-Morgan
posted
an
update
3 months ago
Post
666
retrain-pipelines 0.1.2 finally dropped. It comes with a hot Hugging Face Hub integration. Go check it out. We have 2 articles about it coming up. One already fully written so, be on the lookout !@retrain-pipelines
Also, I'll be volunteering at GOSIM AI Paris 2025. If you're interested in chatting, hmu.
Post
2659
@SmallDoge
SmallTalks(
SmallDoge/SmallTalks) is a synthetic dataset designed for supervised fine-tuning of language models. The dataset covers a variety of conversational content, including daily conversations, tool usage, Python programming, encyclopedia Q&A, exam problem-solving, logical reasoning, and more. Each task is provided in both English and Chinese versions.
hannayukhymenko
posted
an
update
3 months ago
Post
3557
🚀 We are delighted to announce MamayLM, a new state-of-the-art efficient Ukrainian LLM!
📈 MamayLM surpasses similar-sized models in both English and Ukrainian, while matching or overtaking up to 10x larger models.
📊 MamayLM is a 9B model that can run on a single GPU, enabling cost-efficient AI autonomy and adoption across sectors in Ukraine such as education, legal, healthcare, public services and others (e.g., by specializing it to particular use cases). MalayLM is also attractive for organizations wishing to preserve data privacy as it s efficiency allows it to run on a local machine.
🧠 MamayLM is trained on high-quality Ukrainian data and understands Ukrainian language, culture, and history. It is built on top of Google’s Gemma 2 9B model, but uses a number of new advances stemming from INSAIT’s experience in creating BgGPT, a Bulgarian LLM we released last year, now adopted nationwide and profiled several times by Google as a worldwide success case.
🤝 MamayLM is developed in a collaboration between researchers at INSAIT and ETH Zürich and is trained entirely via donations to INSAIT for AI compute resources.
📥 MamayLM is now freely available to download on INSAIT’s HuggingFace in both full and quantized versions. We also publicly release all Ukrainian benchmarks we evaluated on.
📝 Further, we release blog posts in both English and Ukrainian, sharing our approach to creating MamayLM, hoping to drive further improvements by the community.
🌎 The release of LLMs for various languages is part of INSAIT’s mission in ensuring countries can achieve AI autonomy in a cost-efficient, controlled, safe and predictable manner.
MamayLM model and benchmarks:
INSAIT-Institute
Blog (EN): https://huggingface.co/blog/INSAIT-Institute/mamaylm
Blog (UKR): https://huggingface.co/blog/INSAIT-Institute/mamaylm-ukr
📈 MamayLM surpasses similar-sized models in both English and Ukrainian, while matching or overtaking up to 10x larger models.
📊 MamayLM is a 9B model that can run on a single GPU, enabling cost-efficient AI autonomy and adoption across sectors in Ukraine such as education, legal, healthcare, public services and others (e.g., by specializing it to particular use cases). MalayLM is also attractive for organizations wishing to preserve data privacy as it s efficiency allows it to run on a local machine.
🧠 MamayLM is trained on high-quality Ukrainian data and understands Ukrainian language, culture, and history. It is built on top of Google’s Gemma 2 9B model, but uses a number of new advances stemming from INSAIT’s experience in creating BgGPT, a Bulgarian LLM we released last year, now adopted nationwide and profiled several times by Google as a worldwide success case.
🤝 MamayLM is developed in a collaboration between researchers at INSAIT and ETH Zürich and is trained entirely via donations to INSAIT for AI compute resources.
📥 MamayLM is now freely available to download on INSAIT’s HuggingFace in both full and quantized versions. We also publicly release all Ukrainian benchmarks we evaluated on.
📝 Further, we release blog posts in both English and Ukrainian, sharing our approach to creating MamayLM, hoping to drive further improvements by the community.
🌎 The release of LLMs for various languages is part of INSAIT’s mission in ensuring countries can achieve AI autonomy in a cost-efficient, controlled, safe and predictable manner.
MamayLM model and benchmarks:
INSAIT-Institute
Blog (EN): https://huggingface.co/blog/INSAIT-Institute/mamaylm
Blog (UKR): https://huggingface.co/blog/INSAIT-Institute/mamaylm-ukr
- 1 reply
Post
2943
New king of open VLMs: InternVL3 takes Qwen 2.5's crown! 👑
InternVL have been a wildly successful series of model : and the latest iteration has just taken back their crown thanks to their superior, natively multimodal vision training pipeline.
➡️ Most of the vision language models (VLMs) these days are built like Frankenstein : take a good text-only Large Language Model (LLM) backbone, stitch a specific vision transformer (ViT) on top of it. Then the training is sequential 🔢 : 1. Freeze the LLM weights while you train the ViT only to work with the LLM part, then 2. Unfreeze all weights to train all weights in order to work together.
💫 The Shanghai Lab decided to challenge this paradigm and chose this approach that they call "native". For each of their model sizes, they still start from a good LLM (mostly Qwen-2.5 series, did I tell you I'm a huge fan of Qwen? ❤️), and stitch the ViT, but they don't freeze anything : they train all weights together with interleaved text and image understanding data in a single pre-training phase 🎨.
They claim it results in more seamless interactions between modalities. And the results prove them right: they took the crown of top VLMs, at nearly all sizes, from their Qwen-2.5 parents. 👑
InternVL have been a wildly successful series of model : and the latest iteration has just taken back their crown thanks to their superior, natively multimodal vision training pipeline.
➡️ Most of the vision language models (VLMs) these days are built like Frankenstein : take a good text-only Large Language Model (LLM) backbone, stitch a specific vision transformer (ViT) on top of it. Then the training is sequential 🔢 : 1. Freeze the LLM weights while you train the ViT only to work with the LLM part, then 2. Unfreeze all weights to train all weights in order to work together.
💫 The Shanghai Lab decided to challenge this paradigm and chose this approach that they call "native". For each of their model sizes, they still start from a good LLM (mostly Qwen-2.5 series, did I tell you I'm a huge fan of Qwen? ❤️), and stitch the ViT, but they don't freeze anything : they train all weights together with interleaved text and image understanding data in a single pre-training phase 🎨.
They claim it results in more seamless interactions between modalities. And the results prove them right: they took the crown of top VLMs, at nearly all sizes, from their Qwen-2.5 parents. 👑
-
- 2 replies
mrfakename
posted
an
update
3 months ago
Post
2929
Papla P1 from Papla Media is now available on the TTS Arena!
Try out Papla's new ultra-realistic TTS model + compare it with other leading models on the TTS Arena: TTS-AGI/TTS-Arena
Try out Papla's new ultra-realistic TTS model + compare it with other leading models on the TTS Arena: TTS-AGI/TTS-Arena
Post
2425
🚀 DeepSeek R1 moment has come for GUI agents: Rule-based Reinforcement Learning gives better results than SFT with 500x smaller datasets!
Traditionally (by which I mean "in the last few months"), GUI agents have been trained with supervised fine-tuning (SFT). This meant, collecting huge datasets of screen captures from people using computers, and using these to fine-tune your model. 📚
👉 But last week, a new paper introduced UI-R1, applying DeepSeek's R1-style rule-based reinforcement learning (RL) specifically to GUI action prediction tasks.
This is big news: with RL, maybe we could build good agents without the need for huge datasets.
UI-R1 uses a unified reward function that evaluates multiple responses from models, optimizing via policy algorithms like Group Relative Policy Optimization (GRPO).
Specifically, the reward function assesses:
🎯 Action type accuracy: Does the predicted action match the ground truth?
📍 Coordinate accuracy (specifically for clicks): Is the predicted click within the correct bounding box?
📑 Output format: Does the model clearly articulate both its reasoning and final action?
Using just 136 carefully selected mobile tasks—compared to 76,000 tasks for larger models like OS-Atlas—UI-R1 shows significant efficiency and improved performance:
📈 Boosted action prediction accuracy from 76% to 89% on AndroidControl.
🌐 Outperformed larger, SFT-trained models (e.g., OS-Atlas-7B), demonstrating superior results with vastly fewer data points (136 tasks vs. 76K).
🔍 Enhanced adaptability and generalization, excelling even in out-of-domain scenarios.
The paper tests this RL-based method only in low-level GUI tasks. Could it generalize to more complex interactions? 🧐
Read the full paper here 👉 UI-R1: Enhancing Action Prediction of GUI Agents by Reinforcement Learning (2503.21620)
Traditionally (by which I mean "in the last few months"), GUI agents have been trained with supervised fine-tuning (SFT). This meant, collecting huge datasets of screen captures from people using computers, and using these to fine-tune your model. 📚
👉 But last week, a new paper introduced UI-R1, applying DeepSeek's R1-style rule-based reinforcement learning (RL) specifically to GUI action prediction tasks.
This is big news: with RL, maybe we could build good agents without the need for huge datasets.
UI-R1 uses a unified reward function that evaluates multiple responses from models, optimizing via policy algorithms like Group Relative Policy Optimization (GRPO).
Specifically, the reward function assesses:
🎯 Action type accuracy: Does the predicted action match the ground truth?
📍 Coordinate accuracy (specifically for clicks): Is the predicted click within the correct bounding box?
📑 Output format: Does the model clearly articulate both its reasoning and final action?
Using just 136 carefully selected mobile tasks—compared to 76,000 tasks for larger models like OS-Atlas—UI-R1 shows significant efficiency and improved performance:
📈 Boosted action prediction accuracy from 76% to 89% on AndroidControl.
🌐 Outperformed larger, SFT-trained models (e.g., OS-Atlas-7B), demonstrating superior results with vastly fewer data points (136 tasks vs. 76K).
🔍 Enhanced adaptability and generalization, excelling even in out-of-domain scenarios.
The paper tests this RL-based method only in low-level GUI tasks. Could it generalize to more complex interactions? 🧐
Read the full paper here 👉 UI-R1: Enhancing Action Prediction of GUI Agents by Reinforcement Learning (2503.21620)