AIEnergyScore/requests_debug
Viewer
โข
Updated
โข
218
โข
206
โข
1
AI Energy Score
community
AI & ML interests
None defined yet.
Recent Activity
View all activity
AIEnergyScore's activity
sashaย
updated
a
dataset
5 days ago
Post
2390
New launch: See the energy use of chatbot conversations, in real time. =)
jdelavande/chat-ui-energy
Great work from @JulienDelavande !
jdelavande/chat-ui-energy
Great work from @JulienDelavande !
Post
2693
New king of open VLMs: InternVL3 takes Qwen 2.5's crown! ๐
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.
โก๏ธ 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 ๐ข : 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.
๐ซ 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? โค๏ธ), 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 ๐จ.
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. ๐
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.
โก๏ธ 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 ๐ข : 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.
๐ซ 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? โค๏ธ), 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 ๐จ.
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. ๐
- 2 replies
Post
3183
Today in Privacy & AI Tooling - introducing a nifty new tool to examine where data goes in open-source apps on ๐ค
HF Spaces have tons (100Ks!) of cool demos leveraging or examining AI systems - and because most of them are OSS we can see exactly how they handle user data ๐๐
That requires actually reading the code though, which isn't always easy or quick! Good news: code LMs have gotten pretty good at automatic review, so we can offload some of the work - here I'm using Qwen/Qwen2.5-Coder-32B-Instruct to generate reports and it works pretty OK ๐
The app works in three stages:
1. Download all code files
2. Use the Code LM to generate a detailed report pointing to code where data is transferred/(AI-)processed (screen 1)
3. Summarize the app's main functionality and data journeys (screen 2)
4. Build a Privacy TLDR with those inputs
It comes with a bunch of pre-reviewed apps/Spaces, great to see how many process data locally or through (private) HF endpoints ๐ค
Note that this is a POC, lots of exciting work to do to make it more robust, so:
- try it: yjernite/space-privacy
- reach out to collab: yjernite/space-privacy
HF Spaces have tons (100Ks!) of cool demos leveraging or examining AI systems - and because most of them are OSS we can see exactly how they handle user data ๐๐
That requires actually reading the code though, which isn't always easy or quick! Good news: code LMs have gotten pretty good at automatic review, so we can offload some of the work - here I'm using Qwen/Qwen2.5-Coder-32B-Instruct to generate reports and it works pretty OK ๐
The app works in three stages:
1. Download all code files
2. Use the Code LM to generate a detailed report pointing to code where data is transferred/(AI-)processed (screen 1)
3. Summarize the app's main functionality and data journeys (screen 2)
4. Build a Privacy TLDR with those inputs
It comes with a bunch of pre-reviewed apps/Spaces, great to see how many process data locally or through (private) HF endpoints ๐ค
Note that this is a POC, lots of exciting work to do to make it more robust, so:
- try it: yjernite/space-privacy
- reach out to collab: yjernite/space-privacy
Post
2320
๐ 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. ๐
๐ 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:
๐ฏ Action type accuracy: Does the predicted action match the ground truth?
๐ Coordinate accuracy (specifically for clicks): Is the predicted click within the correct bounding box?
๐ Output format: Does the model clearly articulate both its reasoning and final action?
Using just 136 carefully selected mobile tasksโcompared to 76,000 tasks for larger models like OS-AtlasโUI-R1 shows significant efficiency and improved performance:
๐ Boosted action prediction accuracy from 76% to 89% on AndroidControl.
๐ Outperformed larger, SFT-trained models (e.g., OS-Atlas-7B), demonstrating superior results with vastly fewer data points (136 tasks vs. 76K).
๐ 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? ๐ง
Read the full paper here ๐ 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. ๐
๐ 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:
๐ฏ Action type accuracy: Does the predicted action match the ground truth?
๐ Coordinate accuracy (specifically for clicks): Is the predicted click within the correct bounding box?
๐ Output format: Does the model clearly articulate both its reasoning and final action?
Using just 136 carefully selected mobile tasksโcompared to 76,000 tasks for larger models like OS-AtlasโUI-R1 shows significant efficiency and improved performance:
๐ Boosted action prediction accuracy from 76% to 89% on AndroidControl.
๐ Outperformed larger, SFT-trained models (e.g., OS-Atlas-7B), demonstrating superior results with vastly fewer data points (136 tasks vs. 76K).
๐ 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? ๐ง
Read the full paper here ๐ UI-R1: Enhancing Action Prediction of GUI Agents by Reinforcement Learning (2503.21620)
sashaย
authored
a
paper
about 1 month ago
megย
authored
a
paper
about 1 month ago
Post
4952
smolagents now support vLLM! ๐ฅณ
As one of the most popular local inference solutions, the community had been asking us to integrate vLLM: after a heavy refactoring of our LLM classes, we've just released smolagents 1.11.0, with a brand new VLLMModel class.
Go try it and tell us what you think!
https://github.com/huggingface/smolagents/blob/45b2c86857b7f7657daaa74e4d17d347e9e2c4a4/src/smolagents/models.py#L497
As one of the most popular local inference solutions, the community had been asking us to integrate vLLM: after a heavy refactoring of our LLM classes, we've just released smolagents 1.11.0, with a brand new VLLMModel class.
Go try it and tell us what you think!
https://github.com/huggingface/smolagents/blob/45b2c86857b7f7657daaa74e4d17d347e9e2c4a4/src/smolagents/models.py#L497
Post
1062
Our new Agentic leaderboard is now live!๐ฅ
If you ever asked which LLM is best for powering agents, we've just made a leaderboard that ranks them all! Built with @albertvillanova , this ranks LLMs powering a smolagents CodeAgent on subsets of various benchmarks. โ
๐ GPT-4.5 comes on top, even beating reasoning models like DeepSeek-R1 or o1. And Claude-3.7-Sonnet is a close second!
The leaderboard also allows you to show the scores of vanilla LLMs (without any agentic setup) on the same benchmarks: this shows the huge improvements brought by agentic setups. ๐ช
(Note that results will be added manually, so the leaderboard might not always have the latest LLMs)
If you ever asked which LLM is best for powering agents, we've just made a leaderboard that ranks them all! Built with @albertvillanova , this ranks LLMs powering a smolagents CodeAgent on subsets of various benchmarks. โ
๐ GPT-4.5 comes on top, even beating reasoning models like DeepSeek-R1 or o1. And Claude-3.7-Sonnet is a close second!
The leaderboard also allows you to show the scores of vanilla LLMs (without any agentic setup) on the same benchmarks: this shows the huge improvements brought by agentic setups. ๐ช
(Note that results will be added manually, so the leaderboard might not always have the latest LLMs)
- 1 reply
yjerniteย
authored
a
paper
2 months ago
Post
4845
We now have a Deep Research for academia: SurveyX automatically writes academic surveys nearly indistinguishable from human-written ones ๐ฅ
Researchers from Beijing and Shanghai just published the first application of a deep research system to academia: their algorithm, given a question, can give you a survey of all papers on the subject.
To make a research survey, you generally follow two steps, preparation (collect and organize papers) and writing (outline creation, writing, polishing). Researchers followed the same two steps and automated them.
๐ฏ For the preparation part, a key part is find all the important references on the given subject.
Researchers first cast a wide net of all relevant papers. But then finding the really important ones is like distilling knowledge from a haystack of information. To solve this challenge, they built an โAttributeTreeโ object that structures key information from citations. Ablating these AttributeTrees significantly decreased structure and synthesis scores, so they were really useful!
๐ For the writing part, key was to get a synthesis that's both short and true. This is not easy to get with LLMs! So they used methods like LLM-based deduplication to shorten the too verbose listings made by LLMs, and RAG to grab original quotes instead of made-up ones.
As a result, their system outperforms previous approaches by far!
As assessed by LLM-judges, the quality score os SurveyX even approaches this of human experts, with 4.59/5 vs 4.75/5 ๐
I advise you to read the paper, it's a great overview of the kind of assistants that we'll get in the short future! ๐ SurveyX: Academic Survey Automation via Large Language Models (2502.14776)
Their website shows examples of generated surveys ๐ http://www.surveyx.cn/
Researchers from Beijing and Shanghai just published the first application of a deep research system to academia: their algorithm, given a question, can give you a survey of all papers on the subject.
To make a research survey, you generally follow two steps, preparation (collect and organize papers) and writing (outline creation, writing, polishing). Researchers followed the same two steps and automated them.
๐ฏ For the preparation part, a key part is find all the important references on the given subject.
Researchers first cast a wide net of all relevant papers. But then finding the really important ones is like distilling knowledge from a haystack of information. To solve this challenge, they built an โAttributeTreeโ object that structures key information from citations. Ablating these AttributeTrees significantly decreased structure and synthesis scores, so they were really useful!
๐ For the writing part, key was to get a synthesis that's both short and true. This is not easy to get with LLMs! So they used methods like LLM-based deduplication to shorten the too verbose listings made by LLMs, and RAG to grab original quotes instead of made-up ones.
As a result, their system outperforms previous approaches by far!
As assessed by LLM-judges, the quality score os SurveyX even approaches this of human experts, with 4.59/5 vs 4.75/5 ๐
I advise you to read the paper, it's a great overview of the kind of assistants that we'll get in the short future! ๐ SurveyX: Academic Survey Automation via Large Language Models (2502.14776)
Their website shows examples of generated surveys ๐ http://www.surveyx.cn/
bgamazayย
updated
2
Spaces
3 months ago
Post
3102
Less is More for Reasoning (LIMO): a 32B model fine-tuned with 817 examples can beat o1-preview on math reasoning! ๐คฏ
Do we really need o1's huge RL procedure to see reasoning emerge? It seems not.
Researchers from Shanghai Jiaotong University just demonstrated that carefully selected examples can boost math performance in large language models using SFT โno huge datasets or RL procedures needed.
Their procedure allows Qwen2.5-32B-Instruct to jump from 6.5% to 57% on AIME and from 59% to 95% on MATH, while using only 1% of the data in previous approaches.
โก The Less-is-More Reasoning Hypothesis:
โฃ Minimal but precise examples that showcase optimal reasoning patterns matter more than sheer quantity
โฃ Pre-training knowledge plus sufficient computational resources at inference levels up math skills
โก๏ธ Core techniques:
โฃ High-quality reasoning chains with self-verification steps
โฃ 817 handpicked problems that encourage deeper reasoning
โฃ Enough inference-time computation to allow extended reasoning
๐ช Efficiency gains:
โฃ Only 817 examples instead of 100k+
โฃ 40.5% absolute improvement across 10 diverse benchmarks, outperforming models trained on 100x more data
This really challenges the notion that SFT leads to memorization rather than generalization! And opens up reasoning to GPU-poor researchers ๐
Read the full paper here ๐ย LIMO: Less is More for Reasoning (2502.03387)
Do we really need o1's huge RL procedure to see reasoning emerge? It seems not.
Researchers from Shanghai Jiaotong University just demonstrated that carefully selected examples can boost math performance in large language models using SFT โno huge datasets or RL procedures needed.
Their procedure allows Qwen2.5-32B-Instruct to jump from 6.5% to 57% on AIME and from 59% to 95% on MATH, while using only 1% of the data in previous approaches.
โก The Less-is-More Reasoning Hypothesis:
โฃ Minimal but precise examples that showcase optimal reasoning patterns matter more than sheer quantity
โฃ Pre-training knowledge plus sufficient computational resources at inference levels up math skills
โก๏ธ Core techniques:
โฃ High-quality reasoning chains with self-verification steps
โฃ 817 handpicked problems that encourage deeper reasoning
โฃ Enough inference-time computation to allow extended reasoning
๐ช Efficiency gains:
โฃ Only 817 examples instead of 100k+
โฃ 40.5% absolute improvement across 10 diverse benchmarks, outperforming models trained on 100x more data
This really challenges the notion that SFT leads to memorization rather than generalization! And opens up reasoning to GPU-poor researchers ๐
Read the full paper here ๐ย LIMO: Less is More for Reasoning (2502.03387)
How do these numbers compare to the findings of Sasha Luccioni`s paper from last year?
4
#3 opened 3 months ago
by
cisfer
Post
1731
Nice paper comparing the fp8 inference efficiency of Nvidia H100 and Intel Gaudi2:
An Investigation of FP8 Across Accelerators for LLM Inference (2502.01070)
The conclusion is interesting: "Our findings highlight that the Gaudi 2, by leveraging FP8, achieves higher throughput-to-power efficiency during LLM inference"
One aspect of AI hardware accelerators that is often overlooked is how they consume less energy than GPUs. It's nice to see researchers starting carrying out experiments to measure this!
Gaudi3 results soon...
The conclusion is interesting: "Our findings highlight that the Gaudi 2, by leveraging FP8, achieves higher throughput-to-power efficiency during LLM inference"
One aspect of AI hardware accelerators that is often overlooked is how they consume less energy than GPUs. It's nice to see researchers starting carrying out experiments to measure this!
Gaudi3 results soon...
Post
2954
๐๐ฟ๐ฒ๐ฎ๐ ๐ณ๐ฒ๐ฎ๐๐๐ฟ๐ฒ ๐ฎ๐น๐ฒ๐ฟ๐: you can now share agents to the Hub! ๐ฅณ๐ฅณ
And any agent pushed to Hub get a cool Space interface to directly chat with it.
This was a real technical challenge: for instance, serializing tools to export them meant that you needed to get all the source code for a tool, verify that it was standalone (not relying on external variables), and gathering all the packages required to make it run.
Go try it out! ๐ https://github.com/huggingface/smolagents
And any agent pushed to Hub get a cool Space interface to directly chat with it.
This was a real technical challenge: for instance, serializing tools to export them meant that you needed to get all the source code for a tool, verify that it was standalone (not relying on external variables), and gathering all the packages required to make it run.
Go try it out! ๐ https://github.com/huggingface/smolagents
- 2 replies
Post
2554
For those who haven't come across it yet, here's a handy trick to discuss an entire GitHub repo with an LLM:
=> Just replace "github" with "gitingest" in the url, and you get the whole repo as a single string that you can then paste in your LLMs
=> Just replace "github" with "gitingest" in the url, and you get the whole repo as a single string that you can then paste in your LLMs
Post
4869
"๐ฎ๐ฌ๐ฎ๐ฑ ๐๐ถ๐น๐น ๐ฏ๐ฒ ๐๐ต๐ฒ ๐๐ฒ๐ฎ๐ฟ ๐ผ๐ณ ๐๐ ๐ฎ๐ด๐ฒ๐ป๐๐": this statement has often been made, here are numbers to support it.
I've plotted the progress of AI agents on GAIA test set, and it seems they're headed to catch up with the human baseline in early 2026.
And that progress is still driven mostly by the improvement of base LLMs: progress would be even faster with fine-tuned agentic models.
I've plotted the progress of AI agents on GAIA test set, and it seems they're headed to catch up with the human baseline in early 2026.
And that progress is still driven mostly by the improvement of base LLMs: progress would be even faster with fine-tuned agentic models.
bgamazayย
published
a
Space
3 months ago