Update README.md

README.md

@@ -1,81 +1,45 @@
 ---
+license: apache-2.0
 language:
 - en
-license: apache-2.0
 tags:
-- qwen3
+- distillation
 - reasoning
 - math
-- coding
-- autonomous-learning
-- chain-of-thought
-- distillation
-- reinforcement-learning
-base_model: Qwen/Qwen3-0.6B
+- code
+- science
+- gguf
+- spermllm
+- qwen
+- small-language-model
+base_model:
+- Qwen/Qwen3-0.6B-Instruct
 pipeline_tag: text-generation
-datasets:
-- microsoft/orca-math-word-problems-200k
-- meta-math/MetaMathQA
-- theblackcat102/evol-code-alpaca-v1
-- nickrosh/Evol-Instruct-Code-80k-v1
-library_name: transformers
 model-index:
-- name: SpermLLM-S1-Qwen3
-  results:
-  - task:
-      type: text-generation
-    dataset:
-      name: GSM8K
-      type: gsm8k
-    metrics:
-    - type: accuracy
-      value: TBD
-  - task:
-      type: text-generation
-    dataset:
-      name: HumanEval
-      type: openai_humaneval
-    metrics:
-    - type: pass@1
-      value: TBD
+- name: SpermLLM
+  results: []
 ---
 
-<div align="center">
-
-# 🧬 SpermLLM-S1
-
-### *Autonomous Learning Meets Small Language Models*
-
-[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
-[![Model Size](https://img.shields.io/badge/Size-0.6B_params-green.svg)]()
-[![Base Model](https://img.shields.io/badge/Base-Qwen3--0.6B-orange.svg)](https://huggingface.co/Qwen/Qwen3-0.6B)
-[![Distillation](https://img.shields.io/badge/Distilled_from-70B_Teacher-purple.svg)]()
-
-*A 0.6B parameter model that MIGHT punch above it weight class*
+# 🧠 SpermLLM — Distilled Reasoning Model
 
-[🤗 Model](https://huggingface.co/SpermAI/SpermLLM-S1-Qwen3) • [📦 GGUF](https://huggingface.co/SpermAI/SpermLLM-S1-Qwen3-GGUF)
+<p align="center">
+  <img src="https://img.shields.io/badge/Parameters-0.5B-blue" alt="Parameters">
+  <img src="https://img.shields.io/badge/Teacher-Kimi_K2.5_(70B)-green" alt="Teacher">
+  <img src="https://img.shields.io/badge/Method-Auto_Distillation-orange" alt="Method">
+  <img src="https://img.shields.io/badge/Format-GGUF-red" alt="Format">
+  <img src="https://img.shields.io/badge/License-Apache_2.0-purple" alt="License">
+</p>
 
-</div>
+SpermLLM is a compact distilled reasoning model based on **Qwen3-0.6B-Instruct**, designed to improve performance in math, coding, and structured reasoning while remaining lightweight and efficient.
 
----
-
-## 🎯 What Makes SpermLLM Different?
-
-**SpermLLM** isn't just another fine-tuned model. It's trained through **autonomous learning**:
-
-1. 🌐 **Self-Discovers Problems** - Scrapes math and coding challenges from the web
-2. 🧠 **Learns from 120B Teachers** - Gets solutions from GPT-OSS-120B via distillation
-3. 🔄 **Continuous Self-Improvement** - Trains on its failures, gets smarter over time
-4. 🛡️ **Benchmark Decontaminated** - Zero test set leakage (proven via n-gram analysis)
-
-Unlike traditional fine-tuning, SpermLLM **generates its own curriculum** and learns **continuously**.
-
----
+## Training Method
 
-## 🏆 Performance
+The model was fine-tuned on a mixture of curated instruction datasets and further distilled from larger teacher models (Mix of GPT-OSS-120B and Kimi K2.5)
 
-Not yet known. We will test it
+## Training Overview
+- **Base Model**: Qwen3 0.6B Instruct
+- **Training Method**: SFT (Supervised Finetuning) + Distillation
 
-*Note: Benchmarks pending*
+## Notes
 
-**Key Strength:** Step-by-step reasoning (uses `<think>` tags for chain-of-thought)
+SpermLLM is an experimental model, We plan on making this larger and better! Currently no benchmarks but benchmarks will be soon!