AI BriefingSupraLabs Releases Reasoning Dataset with 5 Million Entries
Key point
A Reasoning Corpus dataset with 5 million entries designed to improve the reasoning performance of small language models (SLMs) has been released.
Details
SupraLabs released the reasoning-corpus-4K-5M-v1 dataset on Hugging Face, designed for SFT (Supervised Fine-tuning) of small language models (SLMs).
This dataset consists of a total of 5 million samples, with the following key features:
- Data Composition: Includes user prompts (user), the model's thought process (thought_trace), final answers (assistant), and an integrated ChatML format.
- Training Optimization: All samples are structured within a 5k sequence length, making them optimized for training small models with limited resources.
- Data Fields: Each sample includes metadata such as source (repo_id), token length (tok_len), user input, thought trace, and final answer.
This dataset focuses on helping models perform logical reasoning through step-by-step thinking (Chain-of-Thought).
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Ring-Zero: Emergent Reasoning via Zero RL Scaled to 1 Trillion Parameters
Toss Builds 'Toss Benchmark' to Bridge Gap Between Public Leaderboards and Real-World PerformanceThe Impact of LLM 'Linguistic Illegibility' on Security: Limitations of CoT Monitoring and Proposed Isolation Techniques