Atharva Pandey

The better LLMs get at reasoning, the longer their traces get—thousands of tokens, dozens of tool calls. But in law, medicine, and agentic AI, "usually correct" isn't good enough: answers must be verifiably correct. We built interwhen at @MSFTResearch to make that tractable. And it's now open source. Across benchmarks, plugging interwhen into an LLM yields: ✅ 100% soundness (with full verifiers) 📈 up to 15% accuracy gain ⚡ ~ 1.5× compute cost 🧵

Hello everyone! We just released a small teaser of what we have been working on. Check it out, and hmu if you want to talk more about this space.

Great discussion this evening at the Hidden Layers reading group on State-Space Models. Thanks to @SaiMadhavanG (IIIT-B) and @cloneofatharva (Kairosity Co-founder) for leading us through it with their insights!

So cool...

Excited to share that our paper “STACKFEED”, in collaboration with @namak_kun will be presented at EMNLP 2025! 🎉We explore how to make knowledge bases in RAG systems editable and adaptive instead of staying static and stale.

You are an AI agent... But it's @gautam_jajoo with a spreadsheet and three coffees

RE-IMAGINE Symbolic Benchmark Synthesis for Reasoning Evaluation

@gautam_jajoo Congratulations 👏🎉

full of ups and downs finally fulfilled one dream that my parents always had!

Anyone attending ICML 2025 looking to share accommodation? or have a place for another person?

finally some good dosa in vancouver. credits to @rao2z !