MiroMindAI

@miromind_ai

Building a reasoning-first AI system, provably right for the highest-stakes problems.

miromind.ai

Joined August 2025

193 Following

3,424 Followers

Replies Reposts Media Articles

MiroMindAI
@miromind_ai
May 18
We post-train LLMs for math, for code, for instruction-following. Why not for scientific discovery? 🫎 MOOSE-Star (ICML 2026) : the first scalable SFT recipe for discipline-agnostic scientific hypothesis discovery. github.com/ZonglinY/MOOSE… By @Yang_zy223 & @LidongBing from
GitHub - ZonglinY/MOOSE-Star: [ICML 2026]
From github.com
12
MiroMindAI
@miromind_ai
May 17
Just wrapped our first @aiDotEngineer singapore. Hats off to @agrimsingh @SherryYanJiang @swyx for this incredible event. 3 days at Capitol Singapore, hundreds of builders in the room. A year ago, "more AI Engineers than ML engineers" was a forecast. Today, you could feel it on
6
MiroMindAI
@miromind_ai
May 15
Plug a 300-step deep research agent into your stack—at $1.25/Mtok input. For builders shipping legal analysis, financial risk modeling, or research workflows that need deep reasoning, not just chat. 100 free calls on signup @ platform.miromind.ai/register Built with the only AI that
MiroMindAI Platform - Build Together. Evolve Forever.
From platform.miromind.ai
12
MiroMindAI
@miromind_ai
May 13
Accepted at #ICML2026 — from our team. Hats off to @Yang_zy223 and @LidongBing 🚀🚀 The first scalable recipe for training LLMs in scientific hypothesis discovery — not just prompting them. Brute-force sampling hits a complexity wall but structured search doesn't. This is the
Zonglin Yang
@Yang_zy223
May 13
Can we actually TRAIN LLMs for scientific discovery — or only prompt them to brainstorm? 🧬✨ 🎉 MOOSE-Star → #ICML2026 Most work on LLMs for hypothesis discovery focuses on inference-time agents or feedback-driven refinement. The core generative process — P(hypothesis |
3
MiroMindAI
@miromind_ai
May 11
🚀The MiroThinker API went live! Interactive scaling —up to 300 tool interactions per task × 256K context: 🧠Two engines: ▸ mirothinker-1-7-deepresearch (235B) — GAIA-Val 82.7 · BrowseComp 74.0 · HLE-Text 42.9 ▸ -mini (30B) — BrowseComp-ZH 72.3 (SOTA at 30B) Plus the agent
3

MiroMindAI

MiroMindAI’s posts