TAROT: Test-driven and Capability-adaptive Curriculum Reinforcement Fine-tuning for Code Generation with Large Language Models
Paper • 2602.15449 • Published • 7
AI & ML interests
None defined yet.
Recent Activity
View all activity
chansung
authored a
paper about 2 months ago
chansung
submitted a
paper to Daily Papers about 2 months ago
Post
4732
YAML engineering becomes more and more important than ever from infra provisioning to model training (recipes).
Here, I built a simple editor first for @dstackai , and I will share the live endpoint this week. Let me know what you think about this approach.
Based on this approach, if people think this is useful, I am going to do the same thing for the LLM training recipes for popular frameworks such as Hugging Face open-r1, Axolotl, and so on. Let me hear.
Here, I built a simple editor first for @dstackai , and I will share the live endpoint this week. Let me know what you think about this approach.
Based on this approach, if people think this is useful, I am going to do the same thing for the LLM training recipes for popular frameworks such as Hugging Face open-r1, Axolotl, and so on. Let me hear.
Post
3959
simple guide on the recipe for GRPO on Open-R1 which is built on top of TRL
I think FastAPI wrapper of vLLM with WeightSyncWorker is pretty cool feature. Also, we have many predefined reward functions out of the box!
I think FastAPI wrapper of vLLM with WeightSyncWorker is pretty cool feature. Also, we have many predefined reward functions out of the box!
- 5 replies
Post
2682
Mistral AI Small 3.1 24B is not only commercial free but also the best model in a single GPU deployment.
I packed up all the information you need to know in a single picture. Hope this helps! :)
I packed up all the information you need to know in a single picture. Hope this helps! :)
- 1 reply
Post
1605
Gemma 3 Release in a nutshell
(seems like function calling is not supported whereas the announcement said so)
(seems like function calling is not supported whereas the announcement said so)
Post
3041
Simple Paper Review #5
I briefly reviewed the paper "SFT Memorizes, RL Generalizes," which compares SFT and RL in post-training of LLM/VLM from HKU, UC Berkeley, Google DeepMind, and New York University
The conclusion suggests SFT excels in memorization, while RL is better for generalization. However, since LLM/VLM should benefit humans beyond just generalization, a mix of SFT and RL is advisable. Typically, some SFT is followed by RL to understand prompt formats and enhance generalization through trial and error.
The study focused on one model, Llama-3.2-Vision-11B, using environments like General Points for arithmetic reasoning and V-IRL for spatial reasoning. Training data was used for both SFT and RL, with evaluations on in-distribution and out-of-distribution data to assess memorization and generalization.
I want to apply RL extensively, but it requires building a similar simulation environment. For domain-specific models, significant investment in creating a "playground" for the model is crucial, as the effort will directly influence the outcomes.
https://arxiv.org/abs/2501.17161
I briefly reviewed the paper "SFT Memorizes, RL Generalizes," which compares SFT and RL in post-training of LLM/VLM from HKU, UC Berkeley, Google DeepMind, and New York University
The conclusion suggests SFT excels in memorization, while RL is better for generalization. However, since LLM/VLM should benefit humans beyond just generalization, a mix of SFT and RL is advisable. Typically, some SFT is followed by RL to understand prompt formats and enhance generalization through trial and error.
The study focused on one model, Llama-3.2-Vision-11B, using environments like General Points for arithmetic reasoning and V-IRL for spatial reasoning. Training data was used for both SFT and RL, with evaluations on in-distribution and out-of-distribution data to assess memorization and generalization.
I want to apply RL extensively, but it requires building a similar simulation environment. For domain-specific models, significant investment in creating a "playground" for the model is crucial, as the effort will directly influence the outcomes.
https://arxiv.org/abs/2501.17161
Post
4526
A brief summary of the o3-mini
The OpenAI o3-mini model is a significant improvement over the o1-mini, reaching o1 performance levels. While generally good, its performance isn't universally better than previous models (o1, o1-prev.) or GPT-4o across all benchmarks. This means workflows should be re-evaluated with each model upgrade.
The o3-mini has "low," "medium," and "high" versions, with "low" being the base model used for benchmarking. It's speculated that the higher versions simply involve more processing. A fair comparison with other models like Gemini 2.0 Thinking or DeepSeek-R1 would likely need to use the "low" version and a similar "think more" mechanism.
The system card is recommended reading due to its comprehensive benchmark data.
https://openai.com/index/openai-o3-mini/
The OpenAI o3-mini model is a significant improvement over the o1-mini, reaching o1 performance levels. While generally good, its performance isn't universally better than previous models (o1, o1-prev.) or GPT-4o across all benchmarks. This means workflows should be re-evaluated with each model upgrade.
The o3-mini has "low," "medium," and "high" versions, with "low" being the base model used for benchmarking. It's speculated that the higher versions simply involve more processing. A fair comparison with other models like Gemini 2.0 Thinking or DeepSeek-R1 would likely need to use the "low" version and a similar "think more" mechanism.
The system card is recommended reading due to its comprehensive benchmark data.
https://openai.com/index/openai-o3-mini/
Post
2043
Simple summary on DeepSeek AI's Janus-Pro: A fresh take on multimodal AI!
It builds on its predecessor, Janus, by tweaking the training methodology rather than the model architecture. The result? Improved performance in understanding and generating multimodal data.
Janus-Pro uses a three-stage training strategy, similar to Janus, but with key modifications:
✦ Stage 1 & 2: Focus on separate training for specific objectives, rather than mixing data.
✦ Stage 3: Fine-tuning with a careful balance of multimodal data.
Benchmarks show Janus-Pro holds its own against specialized models like TokenFlow XL and MetaMorph, and other multimodal models like SD3 Medium and DALL-E 3.
The main limitation? Low image resolution (384x384). However, this seems like a strategic choice to focus on establishing a solid "recipe" for multimodal models. Future work will likely leverage this recipe and increased computing power to achieve higher resolutions.
It builds on its predecessor, Janus, by tweaking the training methodology rather than the model architecture. The result? Improved performance in understanding and generating multimodal data.
Janus-Pro uses a three-stage training strategy, similar to Janus, but with key modifications:
✦ Stage 1 & 2: Focus on separate training for specific objectives, rather than mixing data.
✦ Stage 3: Fine-tuning with a careful balance of multimodal data.
Benchmarks show Janus-Pro holds its own against specialized models like TokenFlow XL and MetaMorph, and other multimodal models like SD3 Medium and DALL-E 3.
The main limitation? Low image resolution (384x384). However, this seems like a strategic choice to focus on establishing a solid "recipe" for multimodal models. Future work will likely leverage this recipe and increased computing power to achieve higher resolutions.
Post
1747
New look for AI powered paper reviews from the list by Hugging Face Daily Papers ( managed by the @akhaliq )
Bookmark the webpage along, check comprehensive reviews by Google DeepMind Gemini 1.5, and listen to audio podcast made by the same tech used in NotebookLM.
Link: https://deep-diver.github.io/ai-paper-reviewer/
This is not an official service by Hugging Face. It is just a service developed by an individual developer using his own money :)
Bookmark the webpage along, check comprehensive reviews by Google DeepMind Gemini 1.5, and listen to audio podcast made by the same tech used in NotebookLM.
Link: https://deep-diver.github.io/ai-paper-reviewer/
This is not an official service by Hugging Face. It is just a service developed by an individual developer using his own money :)
Post
2064
Simple summarization of Evolving Deeper LLM Thinking (Google DeepMind)
The process starts by posing a question.
1) The LLM generates initial responses.
2) These generated responses are evaluated according to specific criteria (program-based checker).
3) The LLM critiques the evaluated results.
4) The LLM refines the responses based on the evaluation, critique, and original responses.
The refined response is then fed back into step 2). If it meets the criteria, the process ends. Otherwise, the algorithm generates more responses based on the refined ones (with some being discarded, some remaining, and some responses potentially being merged).
Through this process, it demonstrated excellent performance in complex scheduling problems (travel planning, meeting scheduling, etc.). It's a viable method for finding highly effective solutions in specific scenarios.
However, there are two major drawbacks:
🤔 An excessive number of API calls are required. (While the cost might not be very high, it leads to significant latency.)
🤔 The evaluator is program-based. (This limits its use as a general method. It could potentially be modified/implemented using LLM as Judge, but that would introduce additional API costs for evaluation.)
https://arxiv.org/abs/2501.09891
The process starts by posing a question.
1) The LLM generates initial responses.
2) These generated responses are evaluated according to specific criteria (program-based checker).
3) The LLM critiques the evaluated results.
4) The LLM refines the responses based on the evaluation, critique, and original responses.
The refined response is then fed back into step 2). If it meets the criteria, the process ends. Otherwise, the algorithm generates more responses based on the refined ones (with some being discarded, some remaining, and some responses potentially being merged).
Through this process, it demonstrated excellent performance in complex scheduling problems (travel planning, meeting scheduling, etc.). It's a viable method for finding highly effective solutions in specific scenarios.
However, there are two major drawbacks:
🤔 An excessive number of API calls are required. (While the cost might not be very high, it leads to significant latency.)
🤔 The evaluator is program-based. (This limits its use as a general method. It could potentially be modified/implemented using LLM as Judge, but that would introduce additional API costs for evaluation.)
https://arxiv.org/abs/2501.09891
Post
2109
Simple Summarization on DeepSeek-R1 from DeepSeek AI
The RL stage is very important.
↳ However, it is difficult to create a truly helpful AI for people solely through RL.
↳ So, we applied a learning pipeline consisting of four stages: providing a good starting point, reasoning RL, SFT, and safety RL, and achieved performance comparable to o1.
↳ Simply fine-tuning other open models with the data generated by R1-Zero (distillation) resulted in performance comparable to o1-mini.
Of course, this is just a brief overview and may not be of much help. All models are accessible on Hugging Face, and the paper can be read through the GitHub repository.
Model:
deepseek-ai
Paper: https://github.com/deepseek-ai/DeepSeek-R1
The RL stage is very important.
↳ However, it is difficult to create a truly helpful AI for people solely through RL.
↳ So, we applied a learning pipeline consisting of four stages: providing a good starting point, reasoning RL, SFT, and safety RL, and achieved performance comparable to o1.
↳ Simply fine-tuning other open models with the data generated by R1-Zero (distillation) resulted in performance comparable to o1-mini.
Of course, this is just a brief overview and may not be of much help. All models are accessible on Hugging Face, and the paper can be read through the GitHub repository.
Model:
Paper: https://github.com/deepseek-ai/DeepSeek-R1
- 1 reply
chansung
authored a
paper over 1 year ago
Post
1997
🎙️ Listen to the audio "Podcast" of every single Hugging Face Daily Papers.
Now, "AI Paper Reviewer" project can automatically generates audio podcasts on any papers published on arXiv, and this is integrated into the GitHub Action pipeline. I sounds pretty similar to hashtag#NotebookLM in my opinion.
🎙️ Try out yourself at https://deep-diver.github.io/ai-paper-reviewer/
This audio podcast is powered by Google technologies: 1) Google DeepMind Gemini 1.5 Flash model to generate scripts of a podcast, then 2) Google Cloud Vertex AI's Text to Speech model to synthesize the voice turning the scripts into the natural sounding voices (with latest addition of "Journey" voice style)
"AI Paper Reviewer" is also an open source project. Anyone can use it to build and own a personal blog on any papers of your interests. Hence, checkout the project repository below if you are interested in!
: https://github.com/deep-diver/paper-reviewer
This project is going to support other models including open weights soon for both text-based content generation and voice synthesis for the podcast. The only reason I chose Gemini model is that it offers a "free-tier" which is enough to shape up this projects with non-realtime batch generations. I'm excited to see how others will use this tool to explore the world of AI research, hence feel free to share your feedback and suggestions!
Now, "AI Paper Reviewer" project can automatically generates audio podcasts on any papers published on arXiv, and this is integrated into the GitHub Action pipeline. I sounds pretty similar to hashtag#NotebookLM in my opinion.
🎙️ Try out yourself at https://deep-diver.github.io/ai-paper-reviewer/
This audio podcast is powered by Google technologies: 1) Google DeepMind Gemini 1.5 Flash model to generate scripts of a podcast, then 2) Google Cloud Vertex AI's Text to Speech model to synthesize the voice turning the scripts into the natural sounding voices (with latest addition of "Journey" voice style)
"AI Paper Reviewer" is also an open source project. Anyone can use it to build and own a personal blog on any papers of your interests. Hence, checkout the project repository below if you are interested in!
: https://github.com/deep-diver/paper-reviewer
This project is going to support other models including open weights soon for both text-based content generation and voice synthesis for the podcast. The only reason I chose Gemini model is that it offers a "free-tier" which is enough to shape up this projects with non-realtime batch generations. I'm excited to see how others will use this tool to explore the world of AI research, hence feel free to share your feedback and suggestions!
- 3 replies
Post
4771
Effortlessly stay up-to-date with AI research trends using a new AI tool, "AI Paper Reviewer" !!
It analyzes a list of Hugging Face Daily Papers(w/ @akhaliq ) and turn them into insightful blog posts. This project leverages Gemini models (1.5 Pro, 1.5 Flash, and 1.5 Flash-8B) for content generation and Upstage Document Parse for parsing the layout and contents.
blog link: https://deep-diver.github.io/ai-paper-reviewer/
Also, here is the link of GitHub repository for parsing and generating pipeline. By using this, you can easily build your own GitHub static pages based on any arXiv papers with your own interest!
: https://github.com/deep-diver/paper-reviewer
It analyzes a list of Hugging Face Daily Papers(w/ @akhaliq ) and turn them into insightful blog posts. This project leverages Gemini models (1.5 Pro, 1.5 Flash, and 1.5 Flash-8B) for content generation and Upstage Document Parse for parsing the layout and contents.
blog link: https://deep-diver.github.io/ai-paper-reviewer/
Also, here is the link of GitHub repository for parsing and generating pipeline. By using this, you can easily build your own GitHub static pages based on any arXiv papers with your own interest!
: https://github.com/deep-diver/paper-reviewer
chansung
authored a
paper over 1 year ago
Post
4032
🦙🦙 LLaMA Duo project update
Last time, I gave a brief introduction about LLaMA Duo project with @sayakpaul . It is a simple toolset to aligning sLLM with service LLM with coverage dataset 👉🏻 (https://huggingface.co/posts/chansung/708646454991943).
- coverage dataset is what we believe to be the most important/desired (instruction, response) pairs. In system thinking, each instruction could be an analogy of a function from traditional programming. We make unit tests and measure the coverage % for all the features/functions. Similarly, we need to ensure if our fine-tuned model could handle what % of given instructions from coverage dataset satisfactory (hence coverage dataset).
We have tested it with "Coding" category of data from HuggingFaceH4/no_robots dataset. It has about 300 SFT training data points under Coding category. After fine-tuning Gemma 7B model on that, the result was very poor. LLaMA Duo's evaluation tool gave < 20% of metrics in similarity and preciseness on the test split.
So, we used LLaMA Duo's synthetic data generation tool to generate 60k data points that looks similar to the original dataset. We first created ~10k synthetic data points, then created 50k more based on the synthetic dataset itself.
After fine-tuning Gemma 7B on the 60k synthetic dataset, the evaluation result went up to 80~90% high. Also, when testing out the model in UI, it tends to give good responses.
It is a good showcase to transition from service LLM to sLLM or having a backup sLLM for service LLM failure scenarios. I am going to expand this experiments on all categories of no_robots dataset. It will roughly generate > 100k data points.
Here are some links:
- LLaMA Duo project repo: https://github.com/deep-diver/llamaduo
- 60k Coding synthetic dataset: chansung/merged_ds_coding
- Fine-tuned Gemma 7B model: chansung/coding_llamaduo_60k_v0.2
Last time, I gave a brief introduction about LLaMA Duo project with @sayakpaul . It is a simple toolset to aligning sLLM with service LLM with coverage dataset 👉🏻 (https://huggingface.co/posts/chansung/708646454991943).
- coverage dataset is what we believe to be the most important/desired (instruction, response) pairs. In system thinking, each instruction could be an analogy of a function from traditional programming. We make unit tests and measure the coverage % for all the features/functions. Similarly, we need to ensure if our fine-tuned model could handle what % of given instructions from coverage dataset satisfactory (hence coverage dataset).
We have tested it with "Coding" category of data from HuggingFaceH4/no_robots dataset. It has about 300 SFT training data points under Coding category. After fine-tuning Gemma 7B model on that, the result was very poor. LLaMA Duo's evaluation tool gave < 20% of metrics in similarity and preciseness on the test split.
So, we used LLaMA Duo's synthetic data generation tool to generate 60k data points that looks similar to the original dataset. We first created ~10k synthetic data points, then created 50k more based on the synthetic dataset itself.
After fine-tuning Gemma 7B on the 60k synthetic dataset, the evaluation result went up to 80~90% high. Also, when testing out the model in UI, it tends to give good responses.
It is a good showcase to transition from service LLM to sLLM or having a backup sLLM for service LLM failure scenarios. I am going to expand this experiments on all categories of no_robots dataset. It will roughly generate > 100k data points.
Here are some links:
- LLaMA Duo project repo: https://github.com/deep-diver/llamaduo
- 60k Coding synthetic dataset: chansung/merged_ds_coding
- Fine-tuned Gemma 7B model: chansung/coding_llamaduo_60k_v0.2
Post
4431
💻 Smoothing the Transition from Service LLM to Local LLM
Imagine your go-to LLM service is down, or you need to use it offline – yikes! This project is all about having that "Plan B" ready to go. Here's LLaMA Duo I've been building with @sayakpaul :
✨ Fine-tune a smaller LLM: We used Hugging Face's alignment-handbook to teach a smaller-sized LLM to mimic my favorite large language model. Think of it as that super-smart AI assistant getting a capable understudy.
🤖 Batch Inference: Let's get that fine-tuned LLM working! My scripts generate lots of text like a champ, and we've made sure things run smoothly even with bigger workloads.
🧐 Evaluation: How well is my small LLM doing? We integrated with the Gemini API to use it as an expert judge – it compares my model's work to the original. Talk about a tough critic!
🪄 Synthetic Data Generation: Need to boost that model's performance? Using Gemini's feedback, we can create even more training data, custom-made to make the LLM better.
🧱 Building Blocks: This isn't just a one-time thing – it's a toolkit for all kinds of LLMOps work. Want to change your evaluation metrics? Bring in models trained differently? Absolutely, let's make it happen.
Why this project is awesome:
💪 Reliability: Keep things running no matter what happens to your main LLM source.
🔒 Privacy: Process sensitive information on your own terms.
🗺️ Offline capable: No internet connection? No problem!
🕰️ Version Control: Lock in your favorite LLM's behavior, even if the service model changes.
We'm excited to share the code on GitHub. Curious to see what you all think! 👉🏻 https://github.com/deep-diver/llamaduo
Imagine your go-to LLM service is down, or you need to use it offline – yikes! This project is all about having that "Plan B" ready to go. Here's LLaMA Duo I've been building with @sayakpaul :
✨ Fine-tune a smaller LLM: We used Hugging Face's alignment-handbook to teach a smaller-sized LLM to mimic my favorite large language model. Think of it as that super-smart AI assistant getting a capable understudy.
🤖 Batch Inference: Let's get that fine-tuned LLM working! My scripts generate lots of text like a champ, and we've made sure things run smoothly even with bigger workloads.
🧐 Evaluation: How well is my small LLM doing? We integrated with the Gemini API to use it as an expert judge – it compares my model's work to the original. Talk about a tough critic!
🪄 Synthetic Data Generation: Need to boost that model's performance? Using Gemini's feedback, we can create even more training data, custom-made to make the LLM better.
🧱 Building Blocks: This isn't just a one-time thing – it's a toolkit for all kinds of LLMOps work. Want to change your evaluation metrics? Bring in models trained differently? Absolutely, let's make it happen.
Why this project is awesome:
💪 Reliability: Keep things running no matter what happens to your main LLM source.
🔒 Privacy: Process sensitive information on your own terms.
🗺️ Offline capable: No internet connection? No problem!
🕰️ Version Control: Lock in your favorite LLM's behavior, even if the service model changes.
We'm excited to share the code on GitHub. Curious to see what you all think! 👉🏻 https://github.com/deep-diver/llamaduo
Post
2542
Realize LLM powered idea on Hugging Face Space.
I made Space for you to duplicate, then it comes with Gradio and LLM served by Hugging Face's efficient Text Generation Inference(TGI) framework packed into a single machine.
It provides a sample app code snippet with
Have you ever enjoyed playing with Hugging Chat? Then, you will enjoy writing your own idea with this. Because both are built on top of TGI!
Focus on your app code, and go beyond chat!
chansung/gradio_together_tgi
I made Space for you to duplicate, then it comes with Gradio and LLM served by Hugging Face's efficient Text Generation Inference(TGI) framework packed into a single machine.
It provides a sample app code snippet with
gr.ChatInterface. However, it is not limited to chat usage, but you can leverage the efficiency of TGI for any sort of apps built in Gradio. Have you ever enjoyed playing with Hugging Chat? Then, you will enjoy writing your own idea with this. Because both are built on top of TGI!
Focus on your app code, and go beyond chat!
chansung/gradio_together_tgi
- 2 replies
Post
🎥 🤾 Vid2Persona: talk to person from video clip
A fun project over the last week with @sayakpaul . It has a simple pipeline from extracting traits of video characters to chatting with them.
Under the hood, this project leverages the power of both commercial and open source models. We used Google's Gemini 1.0 Pro Vision model to understand the video content directly, then we used HuggingFaceH4/zephyr-7b-beta model to make conversation!
Try it Hugging Face Space and let us know what you think.
: chansung/vid2persona
The space application is a dedicated implementation for ZeroGPU environment + Hugging Face Inference API with PRO account. If you wish to host it on your own environment, consider duplicate the space or run locally with the project repository
: https://github.com/deep-diver/Vid2Persona
A fun project over the last week with @sayakpaul . It has a simple pipeline from extracting traits of video characters to chatting with them.
Under the hood, this project leverages the power of both commercial and open source models. We used Google's Gemini 1.0 Pro Vision model to understand the video content directly, then we used HuggingFaceH4/zephyr-7b-beta model to make conversation!
Try it Hugging Face Space and let us know what you think.
: chansung/vid2persona
The space application is a dedicated implementation for ZeroGPU environment + Hugging Face Inference API with PRO account. If you wish to host it on your own environment, consider duplicate the space or run locally with the project repository
: https://github.com/deep-diver/Vid2Persona