Hugging Face's logo
LUIS VIZCAYA's picture
1 4

LUIS VIZCAYA PRO

lvizcaya
MAK45's profile picture
Ā·

AI & ML interests

None yet

Recent Activity

repliedto SeaWolf-AI's post 7 days ago
šŸ§¬ Darwin Family: Zero Gradient Steps, GPQA Diamond 88.89% How far can we push LLM reasoning *without* training? Our team at VIDRAFT submitted this paper to Daily Papers yesterday, and it's currently #3. Huge thanks to everyone who upvoted ā€” sharing the core ideas below. šŸ”— Paper: https://huggingface.co/papers/2605.14386 šŸ”— arXiv: https://arxiv.org/abs/2605.14386 šŸ”— Model: https://huggingface.co/FINAL-Bench/Darwin-28B-REASON šŸ”— Model: https://huggingface.co/FINAL-Bench/Darwin-28B-Opus --- TL;DR Darwin Family is a training-free evolutionary merging framework. By recombining the weight spaces of existing LLM checkpoints ā€” with zero gradient-based training ā€” it reaches frontier-level reasoning. - šŸ† Darwin-28B-Opus: GPQA Diamond 88.89% - šŸ’ø Zero gradient steps ā€” not a single B200 or H200 hour needed - šŸ§¬ Consistent gains across 4B ā†’ 35B scale - šŸ”€ Cross-architecture breeding between Transformer and Mamba families - šŸ” Stable recursive multi-generation evolution #Three Core Mechanisms ā‘  14-dim Adaptive Merge Genome ā€” fine-grained recombination at both component level (Attention / FFN / MLP / LayerNorm / Embedding) and block level, expanding the prior evolutionary-merge search space. ā‘” MRI-Trust Fusion ā€” we diagnose each layer's reasoning contribution via an **MRI (Model Reasoning Importance)** signal and fuse it with evolutionary search through a **learnable trust parameter**. Trust the diagnostic too much and search collapses; ignore it and search becomes inefficient ā€” Darwin learns the balance from data. ā‘¢ Architecture Mapper ā€” weight-space breeding across heterogeneous families. Attention Ɨ SSM crossover actually works. Why It Matters > Diagnose latent capabilities already encoded in open checkpoints, > and recombine them ā€” no gradients required. Replies and critiques welcome šŸ™Œ
reacted to SeaWolf-AI's post with šŸ”„ 7 days ago
šŸ§¬ Darwin Family: Zero Gradient Steps, GPQA Diamond 88.89% How far can we push LLM reasoning *without* training? Our team at VIDRAFT submitted this paper to Daily Papers yesterday, and it's currently #3. Huge thanks to everyone who upvoted ā€” sharing the core ideas below. šŸ”— Paper: https://huggingface.co/papers/2605.14386 šŸ”— arXiv: https://arxiv.org/abs/2605.14386 šŸ”— Model: https://huggingface.co/FINAL-Bench/Darwin-28B-REASON šŸ”— Model: https://huggingface.co/FINAL-Bench/Darwin-28B-Opus --- TL;DR Darwin Family is a training-free evolutionary merging framework. By recombining the weight spaces of existing LLM checkpoints ā€” with zero gradient-based training ā€” it reaches frontier-level reasoning. - šŸ† Darwin-28B-Opus: GPQA Diamond 88.89% - šŸ’ø Zero gradient steps ā€” not a single B200 or H200 hour needed - šŸ§¬ Consistent gains across 4B ā†’ 35B scale - šŸ”€ Cross-architecture breeding between Transformer and Mamba families - šŸ” Stable recursive multi-generation evolution #Three Core Mechanisms ā‘  14-dim Adaptive Merge Genome ā€” fine-grained recombination at both component level (Attention / FFN / MLP / LayerNorm / Embedding) and block level, expanding the prior evolutionary-merge search space. ā‘” MRI-Trust Fusion ā€” we diagnose each layer's reasoning contribution via an **MRI (Model Reasoning Importance)** signal and fuse it with evolutionary search through a **learnable trust parameter**. Trust the diagnostic too much and search collapses; ignore it and search becomes inefficient ā€” Darwin learns the balance from data. ā‘¢ Architecture Mapper ā€” weight-space breeding across heterogeneous families. Attention Ɨ SSM crossover actually works. Why It Matters > Diagnose latent capabilities already encoded in open checkpoints, > and recombine them ā€” no gradients required. Replies and critiques welcome šŸ™Œ
updated a Space 8 days ago
lvizcaya/walmart-heb-price-comparator
View all activity

Organizations

Hugging Face Context Course's profile picture ML intern explorers's profile picture