Dipankar Sarkar PRO
dipankarsarkar
AI & ML interests
Building the AI-native stack. Agents as infrastructure, safety as architecture, performance as plumbing. I publish the receipts: papers, datasets, demos.
Recent Activity
upvoted a paper about 4 hours ago
ResearchStudio-Reel: Automate the Last Mile of Research from Paper to Poster, Video, and Blog upvoted a paper about 5 hours ago
SeKV: Resolution-Adaptive KV Cache with Hierarchical Semantic Memory for Long-Context LLM Inference reacted to PhysiQuanty's post with 🔥 about 5 hours ago
🧠 Arithmetic-SLM : A 30M model that manages to compute simple arithmetic better than a 3B model 🚀
https://huggingface.co/WhirlwindAI/Arithmetic-SLM
https://huggingface.co/spaces/WhirlwindAI/arithmetic-slm
🏆 Leaderboard ArithMark-2 🏆
🥇 Qwen/Qwen2.5-Math-1.5B = 82.08%
🥈 WhirlwindAI/Arithmetic-SLM = 78.60% (31.7M Params)
🥉 Qwen/Qwen2.5-3B = 78.44%
Example WhirlwindAI/Arithmetic-SLM =
0.5 * 0.5 = 0.25 ✅
105 + 45 / 8 = 110 ✅
(132 / 12) + (46 - 15) = 42 ✅
(10 + 28) * 3 = 114 ✅
1 * (16 + 28) = 44 ✅
(21 + 27) * (14 - 7) = 336 ❌
```python
leaderboard = """
| Model | Params | Score |
|----------------------------------|--------------|-----------|
| Qwen/Qwen2.5-Math-1.5B | 1.54B | 82.08% |
| WhirlwindAI/Arithmetic-SLM | 31.70M | 78.60% | <=
| Qwen/Qwen2.5-3B | 3.09B | 78.44% |
| Qwen/Qwen2.5-1.5B | 1.54B | 77.72% |
| Qwen/Qwen2.5-Coder-1.5B | 1.54B | 74.88% |
| HuggingFaceTB/SmolLM2-1.7B | 1.71B | 66.12% |
| Qwen/Qwen2.5-0.5B | 494M | 63.04% |
| facebook/MobileLLM-R1-140M-base | 140M | 53.88% |
| SupraLabs/Supra-50M-Base | 52M | 27.12% |
"""
```
Bench =
https://huggingface.co/datasets/AxiomicLabs/ArithMark-2.0
DataSet =
https://huggingface.co/datasets/WhirlwindAI/Arithmetic
By Science AND FOR SCIENCE <3 Organizations
