Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,43 @@
+---
+license: mit
+tags:
+- iknn
+- math
+- agentic
+- deeprcurs
+- IKNN-Rl1-A1
+pretty_name: IKNN-Rl1-Dataset-Math V2 Hard 1000 Examples
+size_categories:
+- 1K<n<10K
+---
+
+# IKNN-Rl1-Dataset-Math — deeprcurs/IKNN-Rl1-A1 — Agentic V2 Hard
+
+**Organization:** deepRcurs Labs  
+**Repo Model:** `deeprcurs/IKNN-Rl1-A1`  
+**Repo Dataset Master:** `deeprcurs/IKNN-Rl1-Dataset-Agentic-V2`  
+**Type:** Math hard — 791 train + 209 val = 1000 total
+
+## Overview
+Split dari dataset agentic V2 10k hard — fokus math — untuk training IKNN-Rl1-A1 researcher agent — correct case SatU1 NoeSA-24 Ntarra-DnA MoE Router PEP RHT PG-KVC ADLP — bukan chatbot biasa.
+
+Dataset V2 10k harder: 2000 logic hard, 2000 reasoning hard, 2000 coding hard, 2000 research hard, 1000 math hard, 1000 science hard — generated by frontier LLM.
+
+## Training V2 Results
+- Steps 500 batch 8 temp 1.0->0.01 2-state {0,3} first 250 then 3-state {0,2,4}
+- Final Loss 0.2019 Val 0.2066 — Improvement from V1 0.4175 (51% reduction)
+- Checkpoint 21MB di model repo deeprcurs/IKNN-Rl1-A1 checkpoints/IKNN-Rl1-A1-150M-agentic-v2.pt — .cache excluded dari snapshot — aman tidak ganggu snapshot
+
+## Files
+- train-math-v2.json 791 examples
+- val-math-v2.json 209 examples
+
+## Link
+- Model: https://huggingface.co/deeprcurs/IKNN-Rl1-A1
+- Dataset Master: https://huggingface.co/datasets/deeprcurs/IKNN-Rl1-Dataset-Agentic-V2
+- Dataset Master All Types: https://huggingface.co/datasets/deeprcurs/IKNN-Rl1-Dataset
+
+## Checkpoint Note
+Checkpoint .pt 21MB disimpan di model repo — tidak mengganggu snapshot workspace karena file asli di .cache/ excluded — upload via HF API langsung — snapshot tetap 1.8M tidy.
+
+## License MIT deepRcurs Labs 2026