FINAL-Bench/Darwin-28B-REASON
Text Generation โข 27B โข Updated โข 202 โข 22
Jinki Jeong PRO
Anserwise
AI & ML interests
None yet
Recent Activity
liked a model 2 days ago
FINAL-Bench/Darwin-28B-REASON liked a Space 3 days ago
FINAL-Bench/Darwin-9B-NEG liked a model 3 days ago
mradermacher/Darwin-36B-Opus-i1-GGUFOrganizations
None yet
liked a
model 2 days ago
liked a
Space 3 days ago
liked 5
models 3 days ago
mradermacher/Darwin-36B-Opus-i1-GGUF
35B โข Updated โข 15.2k โข 5
MCult01/muse-darwin9b
Image-Text-to-Text โข 10B โข Updated โข 80 โข 2
VIDraft/Darwin-9B-KOREA
Text Generation โข 10B โข Updated โข 49 โข 7
VIDraft/Darwin-28B-KOREA
Text Generation โข 28B โข Updated โข 110 โข 7
mudler/Darwin-36B-Opus-APEX-GGUF
35B โข Updated โข 4.55k โข 15
reacted to SeaWolf-AI's post with ๐๐ค๐ฅ 5 days ago
Post
5326
๐งฌ 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 โ sharing the core ideas below.
๐ Paper: Darwin Family: MRI-Trust-Weighted Evolutionary Merging for Training-Free Scaling of Language-Model Reasoning (2605.14386)
๐ arXiv: https://arxiv.org/abs/2605.14386
๐ Model: FINAL-Bench/Darwin-28B-REASON
๐ 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 โ with zero
gradient-based training โ it reaches frontier-level reasoning.
- ๐ Darwin-28B-Opus: GPQA Diamond 88.89%
- ๐ธ Zero gradient steps โ not a single B200 or H200 hour needed
- ๐งฌ Consistent gains across 4B โ 35B scale
- ๐ Cross-architecture breeding between Transformer and Mamba families
- ๐ Stable recursive multi-generation evolution
#Three Core Mechanisms
โ 14-dim Adaptive Merge Genome โ fine-grained recombination at both
component level (Attention / FFN / MLP / LayerNorm / Embedding) and block
level, expanding the prior evolutionary-merge search space.
โก MRI-Trust Fusion โ 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 โ Darwin learns the balance from data.
โข Architecture Mapper โ weight-space breeding across heterogeneous
families. Attention ร SSM crossover actually works.
Why It Matters
> Diagnose latent capabilities already encoded in open checkpoints,
> and recombine them โ no gradients required.
Replies and critiques welcome ๐
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 โ sharing the core ideas below.
๐ Paper: Darwin Family: MRI-Trust-Weighted Evolutionary Merging for Training-Free Scaling of Language-Model Reasoning (2605.14386)
๐ arXiv: https://arxiv.org/abs/2605.14386
๐ Model: FINAL-Bench/Darwin-28B-REASON
๐ 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 โ with zero
gradient-based training โ it reaches frontier-level reasoning.
- ๐ Darwin-28B-Opus: GPQA Diamond 88.89%
- ๐ธ Zero gradient steps โ not a single B200 or H200 hour needed
- ๐งฌ Consistent gains across 4B โ 35B scale
- ๐ Cross-architecture breeding between Transformer and Mamba families
- ๐ Stable recursive multi-generation evolution
#Three Core Mechanisms
โ 14-dim Adaptive Merge Genome โ fine-grained recombination at both
component level (Attention / FFN / MLP / LayerNorm / Embedding) and block
level, expanding the prior evolutionary-merge search space.
โก MRI-Trust Fusion โ 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 โ Darwin learns the balance from data.
โข Architecture Mapper โ weight-space breeding across heterogeneous
families. Attention ร SSM crossover actually works.
Why It Matters
> Diagnose latent capabilities already encoded in open checkpoints,
> and recombine them โ no gradients required.
Replies and critiques welcome ๐
- 3 replies
upvoted an article 5 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
โข โข 17 upvoted a paper 5 days ago
liked 8
models 9 days ago
DevQuasar/FINAL-Bench.Darwin-36B-Opus-GGUF
Text Generation โข 35B โข Updated โข 693 โข 6
mradermacher/Darwin-9B-NEG-i1-GGUF
9B โข Updated โข 5.65k โข 7
wahidmounir/Darwin-36B-Opus-GGUF
Text Generation โข 35B โข Updated โข 362 โข 6
mradermacher/Darwin-28B-Opus-GGUF
27B โข Updated โข 1.68k โข 7
mradermacher/Darwin-28B-Opus-i1-GGUF
27B โข Updated โข 5.71k โข 7
fattis/Darwin-2B-Opus-heretic-GGUF
Text Generation โข 2B โข Updated โข 763 โข 6
addansee2/Darwin-2B-Opus-heretic
Text Generation โข 2B โข Updated โข 130 โข 6
Rikunarita-ORG/Darwin-4B-Genesis-i-Q5_K_M-GGUF_Fixed
Text Generation โข 8B โข Updated โข 294 โข 7