akhaliq
Abstract
We present phi-4, a 14-billion parameter language model developed with a training recipe that is centrally focused on data quality. Unlike most language models, where pre-training is based primarily on organic data sources such as web content or code, phi-4 strategically incorporates synthetic data throughout the training process. While previous models in the Phi family largely distill the capabilities of a teacher model (specifically GPT-4), phi-4 substantially surpasses its teacher model on STEM-focused QA capabilities, giving evidence that our data-generation and post-training techniques go beyond distillation. Despite minimal changes to the phi-3 architecture, phi-4 achieves strong performance relative to its size -- especially on reasoning-focused benchmarks -- due to improved data, training curriculum, and innovations in the post-training scheme.
Community
so its no longer "tiny llms that punch above their weight", its just "small" 14B models?
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- Tulu 3: Pushing Frontiers in Open Language Model Post-Training (2024)
- PerfCodeGen: Improving Performance of LLM Generated Code with Execution Feedback (2024)
- Training and Evaluating Language Models with Template-based Data Generation (2024)
- Fox-1 Technical Report (2024)
- ALMA: Alignment with Minimal Annotation (2024)
- Training Data for Large Language Model (2024)
- TransformLLM: Adapting Large Language Models via LLM-Transformed Reading Comprehension Text (2024)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
Thanks for thorough description of various synthetic pipelines!
I have a question about filtering QA pairs. How to apply majority voting to LLM answers?
When the answer is an option it's straightforward, but for open question it won't work.
good question, plurality sampling is mainly beneficial in the scope of higher-reasoning math/science questions and so you can often use another LLM agent to extract some final answer in a specific format, find the majority and fairly compare.
Models citing this paper 1
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper