Anserwise/AWAXIS-KR-31B
Image-Text-to-Text • 52B • Updated • 52 • 2
Jinki Jeong PRO
Anserwise
AI & ML interests
None yet
Recent Activity
liked a model about 8 hours ago
Anserwise/AWAXIS-KR-31B liked a model about 16 hours ago
Anserwise/AWAXIS-Think-31B updated a model about 16 hours ago
Anserwise/AWAXIS-KR-31BOrganizations
None yet
liked a
model about 8 hours ago
liked a
model about 16 hours ago
updated a
model about 16 hours ago
published a
model about 17 hours ago
updated a
model about 18 hours ago
published a
model about 18 hours ago
liked a
model 3 days ago
liked a
Space 4 days ago
liked 5
models 4 days ago
mradermacher/Darwin-36B-Opus-i1-GGUF
35B • Updated • 15.5k • 6
MCult01/muse-darwin9b
Image-Text-to-Text • 10B • Updated • 82 • 2
VIDraft/Darwin-9B-KOREA
Text Generation • 10B • Updated • 52 • 7
VIDraft/Darwin-28B-KOREA
Text Generation • 28B • Updated • 117 • 7
mudler/Darwin-36B-Opus-APEX-GGUF
35B • Updated • 4.89k • 16
reacted to SeaWolf-AI's post with 👍🤗🔥 6 days ago
Post
5339
🧬 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 — sharing the core ideas below.
🔗 Paper: Darwin Family: MRI-Trust-Weighted Evolutionary Merging for Training-Free Scaling of Language-Model Reasoning (2605.14386)
🔗 arXiv: https://arxiv.org/abs/2605.14386
🔗 Model: FINAL-Bench/Darwin-28B-REASON
🔗 Model: 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 — with zero
gradient-based training — it reaches frontier-level reasoning.
- 🏆 Darwin-28B-Opus: GPQA Diamond 88.89%
- 💸 Zero gradient steps — not a single B200 or H200 hour needed
- 🧬 Consistent gains across 4B → 35B scale
- 🔀 Cross-architecture breeding between Transformer and Mamba families
- 🔁 Stable recursive multi-generation evolution
#Three Core Mechanisms
① 14-dim Adaptive Merge Genome — fine-grained recombination at both
component level (Attention / FFN / MLP / LayerNorm / Embedding) and block
level, expanding the prior evolutionary-merge search space.
② MRI-Trust Fusion — 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 — Darwin learns the balance from data.
③ Architecture Mapper — weight-space breeding across heterogeneous
families. Attention × SSM crossover actually works.
Why It Matters
> Diagnose latent capabilities already encoded in open checkpoints,
> and recombine them — no gradients required.
Replies and critiques welcome 🙌
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 — sharing the core ideas below.
🔗 Paper: Darwin Family: MRI-Trust-Weighted Evolutionary Merging for Training-Free Scaling of Language-Model Reasoning (2605.14386)
🔗 arXiv: https://arxiv.org/abs/2605.14386
🔗 Model: FINAL-Bench/Darwin-28B-REASON
🔗 Model: 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 — with zero
gradient-based training — it reaches frontier-level reasoning.
- 🏆 Darwin-28B-Opus: GPQA Diamond 88.89%
- 💸 Zero gradient steps — not a single B200 or H200 hour needed
- 🧬 Consistent gains across 4B → 35B scale
- 🔀 Cross-architecture breeding between Transformer and Mamba families
- 🔁 Stable recursive multi-generation evolution
#Three Core Mechanisms
① 14-dim Adaptive Merge Genome — fine-grained recombination at both
component level (Attention / FFN / MLP / LayerNorm / Embedding) and block
level, expanding the prior evolutionary-merge search space.
② MRI-Trust Fusion — 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 — Darwin learns the balance from data.
③ Architecture Mapper — weight-space breeding across heterogeneous
families. Attention × SSM crossover actually works.
Why It Matters
> Diagnose latent capabilities already encoded in open checkpoints,
> and recombine them — no gradients required.
Replies and critiques welcome 🙌
- 3 replies
upvoted an article 6 days ago
Article
Training-Free Reasoning at 88.89% on GPQA Diamond: How Darwin Family Hit Frontier Scores Without a Single Gradient Step
FINAL-Bench
• • 17 upvoted a paper 6 days ago
liked 2
models 9 days ago
