Update README.md

README.md

@@ -2,6 +2,7 @@
 license: apache-2.0
 base_model:
   - Qwen/Qwen3.5-9B
+  - Jackrong/Qwen3.5-9B-Claude-4.6-Opus-Reasoning-Distilled
 tags:
   - merge
   - evolutionary-merge
@@ -57,29 +58,113 @@ model-index:
 
 # Darwin-9B-Opus
 
-*"Compact reasoning powerhouse — 9B parameters, graduate-level intelligence."*
-
 <p align="center">
-  <img src="info.png" alt="Darwin-35B-A3B-Opus" width="100%">
+  <a href="https://huggingface.co/FINAL-Bench/Darwin-9B-Opus"><img src="https://img.shields.io/badge/Model-Darwin--9B--Opus-blue?style=for-the-badge" alt="Model"></a>
+  <a href="https://huggingface.co/spaces/FINAL-Bench/Darwin-9B-Opus"><img src="https://img.shields.io/badge/Space-9B_Live_Demo-purple?style=for-the-badge" alt="Space"></a>
+  <a href="https://huggingface.co/FINAL-Bench/Darwin-35B-A3B-Opus"><img src="https://img.shields.io/badge/Model-Darwin--35B--A3B--Opus-blue?style=for-the-badge" alt="35B Model"></a>
+  <a href="https://huggingface.co/spaces/FINAL-Bench/Darwin-35B-A3B-Opus"><img src="https://img.shields.io/badge/Space-35B_Live_Demo-purple?style=for-the-badge" alt="35B Space"></a>
+  <a href="https://huggingface.co/spaces/FINAL-Bench/Leaderboard"><img src="https://img.shields.io/badge/FINAL_Bench-Leaderboard-green?style=for-the-badge" alt="FINAL Bench"></a>
+  <a href="https://huggingface.co/spaces/FINAL-Bench/all-bench-leaderboard"><img src="https://img.shields.io/badge/ALL_Bench-Leaderboard-orange?style=for-the-badge" alt="ALL Bench"></a>
 </p>
 
-
 <p align="center">
-  <a href="https://huggingface.co/FINAL-Bench/Darwin-9B-Opus"><img src="https://img.shields.io/badge/🧬_Model-Darwin--9B--Opus-blue?style=for-the-badge" alt="Model"></a>
-  <a href="https://huggingface.co/spaces/FINAL-Bench/Darwin-9B-Opus"><img src="https://img.shields.io/badge/🚀_Space-9B_Live_Demo-purple?style=for-the-badge" alt="Space"></a>
-  <a href="https://huggingface.co/FINAL-Bench/Darwin-35B-A3B-Opus"><img src="https://img.shields.io/badge/🧬_Model-Darwin--35B--A3B--Opus-blue?style=for-the-badge" alt="35B Model"></a>
-  <a href="https://huggingface.co/spaces/FINAL-Bench/Darwin-35B-A3B-Opus"><img src="https://img.shields.io/badge/🚀_Space-35B_Live_Demo-purple?style=for-the-badge" alt="35B Space"></a>
-  <a href="https://huggingface.co/spaces/FINAL-Bench/Leaderboard"><img src="https://img.shields.io/badge/🏆_FINAL_Bench-Leaderboard-green?style=for-the-badge" alt="FINAL Bench"></a>
-  <a href="https://huggingface.co/spaces/FINAL-Bench/all-bench-leaderboard"><img src="https://img.shields.io/badge/📊_ALL_Bench-Leaderboard-orange?style=for-the-badge" alt="ALL Bench"></a>
+  <img src="info.png" alt="Darwin-9B-Opus" width="100%">
 </p>
- 
-> **Qwen3.5 Dense 9B** | Reasoning | Chain-of-Thought | 131K Context | 201 Languages | BF16 | Apache 2.0
+
+> Qwen3.5 Dense 9B | Reasoning | Chain-of-Thought | 131K Context | 201 Languages | BF16 | Apache 2.0
+
+---
+
+## Technical Definitions
+
+| Term | Definition | Measurement |
+|---|---|---|
+| Model MRI | Layer-level profiling of tensor health indicators | L2 norm, Shannon entropy, std per tensor across all layers |
+| LayerMRI.compare_layers | Per-tensor A vs B quality comparison yielding optimal ratio_b | score = entropy * 0.5 + std * 0.3 + clamp(norm, 100) * 0.002 per model; ratio_b = score_b / (score_a + score_b) |
+| MRI-Guided Merge | Per-tensor merge ratios derived from parent diagnostics (70% MRI + 30% genome) | final_ratio = mri_ratio * 0.7 + genome_ratio * 0.3 |
+| DARE-TIES | Merge algorithm: random binary mask on delta, then weighted addition | merged = A + (B - A) * random_mask(density) * ratio |
+| Transplant A / B | When MRI ratio falls below 0.05 or above 0.95, one parent is used entirely | No interpolation — direct tensor copy |
+| Evolutionary Search | CMA-ES population evolution over genome space (ratio, attn, ffn, embed, density_a, density_b) | Phase 1: 200 steps heuristic proxy, Phase 2: 10 steps real benchmark |
 
 ---
 
 ## Overview
 
-Darwin-9B-Opus is a **9B dense parameter** reasoning model created using **Darwin V5**, an evolutionary merge engine with Model MRI integration. Built on the Qwen3.5-9B architecture, it inherits structured step-by-step reasoning capabilities through Claude 4.6 Opus distillation while maintaining the full multilingual and long-context capabilities of the base model.
+Darwin-9B-Opus is a 9B dense parameter reasoning model created using Darwin V5. Both parent models share the identical Qwen3.5-9B architecture — the Mother is a LoRA SFT on the same base, not a different architecture.
+
+| Role | Model | Training |
+|---|---|---|
+| Father | [Qwen/Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B) | Original pre-training + RLHF |
+| Mother | [Jackrong/Qwen3.5-9B-Claude-4.6-Opus-Reasoning-Distilled](https://huggingface.co/Jackrong/Qwen3.5-9B-Claude-4.6-Opus-Reasoning-Distilled) | LoRA SFT with text-only Claude 4.6 Opus reasoning chains |
+
+---
+
+## How Darwin V5 Works
+
+Darwin V5 does not use mergekit or any external merge library. It implements DARE-TIES merge directly via PyTorch tensor operations, with MRI-guided per-layer ratios. The algorithm is inspired by the DARE-TIES method but re-implemented from scratch to support per-tensor diagnostic-guided ratios.
+
+### Merge Implementation (actual code logic)
+
+```python
+# For each tensor pair (A, B) across all safetensor shards:
+ta = model_a[key]       # Father tensor
+tb = model_b[key]       # Mother tensor
+
+# 1. MRI diagnoses both tensors
+diag_a = LayerMRI.diagnose_tensor(ta)  # {norm, entropy, std}
+diag_b = LayerMRI.diagnose_tensor(tb)  # {norm, entropy, std}
+
+# 2. Quality score comparison determines ratio_b
+score_a = diag_a["entropy"] * 0.5 + diag_a["std"] * 0.3 + min(diag_a["norm"], 100) * 0.002
+score_b = diag_b["entropy"] * 0.5 + diag_b["std"] * 0.3 + min(diag_b["norm"], 100) * 0.002
+mri_ratio = score_b / (score_a + score_b)  # Higher = Mother is better
+
+# 3. Final ratio = MRI 70% + evolutionary genome 30%
+final_ratio = mri_ratio * 0.7 + genome_type_ratio * 0.3
+
+# 4. DARE-TIES merge with per-tensor ratio
+mask = torch.rand_like(tb) < density_b
+delta = (tb - ta) * mask
+merged = (ta + delta * final_ratio).bfloat16()
+```
+
+### Pipeline
+
+```
+Phase 0: Model MRI
+  For every tensor in both parents, measure:
+    - L2 norm (layer energy)
+    - Shannon entropy (weight distribution uniformity)
+    - Standard deviation (activation spread)
+  Compare A vs B quality scores -> per-tensor ratio prescription
+
+Phase 1: Evolutionary Search (200 steps, heuristic proxy)
+  Population of 20 genomes (ratio, attn, ffn, embed, density_a, density_b)
+  Fitness: heuristic score based on genome balance + differentiation
+  Selection -> SLERP crossover -> Gaussian mutation
+
+Phase 2: Real Merge + Benchmark (10 steps)
+  Top genomes from Phase 1 undergo actual tensor merge
+  Each merge: MRI prescription (70%) + genome ratio (30%)
+  Fitness: real benchmark score (ARC-Challenge)
+  Best model selected and auto-uploaded
+
+Phase 3: Health Check
+  Layer-by-layer importance comparison: child vs both parents
+  Detect interference (child >> parents) or function loss (parents >> child)
+```
+
+### What Makes This Different from Standard Merging
+
+| Capability | Standard DARE-TIES | Darwin V5 |
+|---|---|---|
+| Implementation | mergekit library call | Direct PyTorch tensor operations |
+| Ratio selection | Uniform ratio across all tensors | Per-tensor ratio from MRI diagnosis |
+| Pre-merge analysis | None | Tensor-level norm/entropy/std profiling |
+| Ratio determination | Human-set or grid search | MRI 70% + evolutionary genome 30% |
+| Post-merge validation | Benchmark score only | Layer-by-layer child vs parents comparison |
+| Transplant support | No | ratio < 0.05 -> use A entirely, ratio > 0.95 -> use B entirely |
+| Failure diagnosis | "Score went down" | Per-tensor quality delta identifies problematic layers |
 
 ---
 
@@ -87,7 +172,7 @@ Darwin-9B-Opus is a **9B dense parameter** reasoning model created using **Darwi
 
 | | |
 |---|---|
-| Architecture | Qwen3.5 Dense |
+| Architecture | Qwen3.5 Dense (Gated DeltaNet hybrid) |
 | Total Parameters | 9B |
 | Precision | BF16 |
 | Context Length | 131,072 native |
@@ -102,10 +187,9 @@ Darwin-9B-Opus is a **9B dense parameter** reasoning model created using **Darwi
 | Setup | VRAM | Status |
 |---|---|---|
 | BF16 Full Precision | ~20 GB | |
-| NVIDIA A10G 24GB | 24 GB | ✅ Comfortable |
-| NVIDIA RTX 4090 24GB | 24 GB | ✅ Comfortable |
-| NVIDIA A100 40GB | 40 GB | ✅ Very comfortable |
-| NVIDIA T4 16GB | 16 GB | ⚠️ Requires quantization |
+| NVIDIA RTX 4090 24GB | 24 GB | Comfortable |
+| NVIDIA A100 40GB | 40 GB | Very comfortable |
+| NVIDIA T4 16GB | 16 GB | Requires quantization |
 
 ---
 
@@ -128,7 +212,7 @@ model = AutoModelForCausalLM.from_pretrained(
     trust_remote_code=True,
 )
 
-messages = [{"role": "user", "content": "Prove that √2 is irrational."}]
+messages = [{"role": "user", "content": "Prove that sqrt(2) is irrational."}]
 text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
 inputs = tokenizer(text, return_tensors="pt").to(model.device)
 outputs = model.generate(**inputs, max_new_tokens=4096)
@@ -156,28 +240,26 @@ vllm serve FINAL-Bench/Darwin-9B-Opus \
 
 ---
 
-## What Makes Darwin Special?
+## Evolution Details
 
-Darwin-9B-Opus was created using **Darwin V5**, an evolutionary merge engine with Model MRI integration.
+| | |
+|---|---|
+| Engine | Darwin V5 (Evolutionary Merge + Layer-Level Diagnostics) |
+| Merge Method | DARE-TIES (direct PyTorch implementation, no external library) |
+| MRI Integration | Per-tensor diagnosis: norm, entropy, std -> ratio prescription |
+| Ratio Formula | final_ratio = mri_ratio * 0.7 + genome_ratio * 0.3 |
+| Evolution | Phase 1: 200 steps proxy + Phase 2: 10 steps real benchmark |
+| Best Score | 0.8508 (ARC-Challenge) |
+| Infrastructure | 4 x NVIDIA H100 NVL (100GB each) |
 
-### Darwin V5 Pipeline
+---
 
-```
-[Phase 0] Model MRI — Profile both parents layer by layer
-    ↓  Measure: layer importance, probe cosine distance
-    ↓
-[Phase 1] MRI-Guided Evolution — Diagnostic-informed initial genome
-    ↓  Not random, but "informed by profiling results"
-    ↓
-[Phase 2] mergekit real merge + benchmark fitness selection
-    ↓  Faster convergence in MRI-narrowed search space
-    ↓
-[Phase 3] MRI Health Check — Profile the child model
-    ↓  Detect interference, function loss
-    ↓  Prescribe layer-specific ratio adjustments
-    ↓
-[Final] Darwin-9B-Opus
-```
+## Acknowledgements
+
+- Korean Government — GPU Support Program research grant
+- [Qwen Team](https://huggingface.co/Qwen) — Qwen3.5 base architecture
+- [Jackrong](https://huggingface.co/Jackrong) — Claude 4.6 Opus Reasoning Distilled model
+- DARE-TIES algorithm — [Yadav et al., 2023](https://arxiv.org/abs/2311.03099) (re-implemented, not library-dependent)
 
 ---
 
@@ -185,35 +267,20 @@ Darwin-9B-Opus was created using **Darwin V5**, an evolutionary merge engine wit
 
 | | |
 |---|---|
-| Developer | **VIDRAFT** |
-| Engine | Darwin V5 (Evolutionary Merge + Model MRI) |
-| Merge Backend | mergekit (DARE-TIES) |
+| Developer | VIDRAFT |
+| Engine | Darwin V5 |
 | Base Architecture | Qwen3.5-9B |
 
 ---
 
-## Acknowledgements
-
-- **Korean Government** — GPU Support Program research grant
-- [Qwen Team](https://huggingface.co/Qwen) — Qwen3.5 base architecture
-- [mergekit](https://github.com/arcee-ai/mergekit) — Merge backend infrastructure
-
----
-
 ## Citation
 
 ```bibtex
 @misc{vidraft_darwin_9b_opus,
-  title        = {Darwin-9B-Opus: Compact Reasoning Model via Diagnostic-Guided Evolutionary Merge},
+  title        = {Darwin-9B-Opus: Diagnostic-Guided Evolutionary Merge},
   author       = {VIDRAFT},
   year         = {2026},
   publisher    = {Hugging Face},
   howpublished = {\url{https://huggingface.co/FINAL-Bench/Darwin-9B-Opus}}
 }
-```
-
----
-
-## Contact
-
-📧 **kkms1116@koreacu.ac.kr**
+```