Update README.md

README.md

@@ -1,12 +1,4 @@
-# DMind-3-nano: Privacy-First On-Device Crypto Intent Recognition
-
-> Inference stays on your device. Standardized function calling for wallets, DEXs, and agents. Built on `google/functiongemma-270m-it`.
-
-**Repo purpose:** host the open-source training/eval pipeline and release artifacts. Place your exported model under `./model` before pushing to Hugging Face.
-
-## HF Card Metadata
-
-```
+---
 language:
 - en
 - zh
@@ -27,7 +19,21 @@ tags:
 - standard-protocol
 library_name: transformers
 pipeline_tag: text-generation
-```
+---
+
+# DMind-3-nano: Privacy-First On-Device Crypto Intent Recognition
+
+> Inference stays on your device. Standardized function calling for wallets, DEXs, and agents. Built on `google/functiongemma-270m-it`.
+
+## Model Description
+
+DMind-3-nano is a small, edge-optimized language model fine-tuned for **crypto wallet and DEX intent recognition** using standardized function-calling protocols. It is designed to run **entirely on-device**, enabling privacy-preserving, low-latency intent parsing for Web3 wallets and local agents.
+
+This repository hosts the **open-source training and evaluation pipeline** as well as the released model artifacts.
+
+**Repo purpose:** host the open-source training/eval pipeline and release artifacts.
+
+
 
 ## Highlights
 
@@ -152,24 +158,16 @@ Model: <start_function_call>call:SEARCH_TOKEN{symbol:"SOL",chain:"solana"}<end_f
 
 ## Performance Snapshot
 
-- Function recognition: ~96.8% on validated set
-- Argument extraction: ~94.2%
-- Protocol adherence: SEARCH_TOKEN 98.1%, EXECUTE_SWAP 95.3%
-- Multi-turn success: ~92.7%
+- Function recognition: ~98.8% on validated set
+- Argument extraction: ~97.4%
+- Protocol adherence: SEARCH_TOKEN 98.5%, EXECUTE_SWAP 97.3%
+- Multi-turn success: ~93.7%
 
 Scope: tokens/chains listed in **Model Overview**; outside that set may be lower.
 
-## Deployment Notes
-
-- On-device: convert to ONNX/CoreML/TFLite for mobile/hardware wallets; apply INT8 quant for ~70MB.
-- CPU-only: expect sub-500ms on modern CPUs; GPUs give faster.
-- Keep Chinese benchmark samples intact (they are intentional test cases).
-
 ## License & Governance
 
 - Code: MIT (`LICENSE`)
 - Model card intent: Apache-2.0 (as in metadata above)
 - Protocol specs (SEARCH_TOKEN / EXECUTE_SWAP): public domain for maximal adoption
 - Contributions are welcome via issues/PRs.
-
-Issues/PRs welcome. When publishing to Hugging Face, ensure `./model` contains your final weights/tokenizer. Replace `YOUR_ORG/DMind-3-nano`, badges, and links with your namespace before release.