📺 Xiaomi Is Livestreaming a Production RL Run — Burn Rate Included @XiaomiMiMo is training its MiMo-V2.6 in public — not a demo, a live post-training dashboard where every step's reward, loss and benchmark movement is watchable in real time. One day in, the meter reads roughly $1.2M spent and 60B tokens consumed. As Zhihu contributor Kitt在进化 puts it: this isn't a training site, it's a money-burning site — about ¥4,000 a minute. His bigger point: this is rare, valuable data. Almost nobody publishes what frontier-scale RL actually costs. 1️⃣ What the dashboard actually shows Every training step exposes the internals most labs keep private. His reading guide: 🔹 Loss family: training loss, entropy (are answers diversifying or collapsing into one mode), gradient norm, and train/inference KL divergence. 🔹 Reward: a mix of test-case scoring and rubric-based reward. 🔹 dynsam (dynamic sampling): avg@n performance over repeated tries, plus pass-rate buckets — including samples never solved and always solved. 🔹 DeepSWE v1.1 as the headline probe (mini-swe-agent, avg@3): Pro at 63.7, Flash at 60.7 at the time of writing. 2️⃣ The cost ledger, out in the open From the dashboard he extracts the numbers the industry usually guesses at: 🔹 MiMo-V2.6 Pro: roughly $36 per million tokens of training. 🔹 MiMo-V2.6 Flash: roughly $8 per million tokens. 🔹 His rule of thumb: equivalent inference runs 30-50x cheaper than these training figures. 3️⃣ What's actually being trained The sample distribution is also public: about 1,500 prompts per step, ~70% of them coding tasks. The remaining third is split across general, visual, cybersecurity and chat tasks — which makes this, in effect, an agentic-coding-centric RL run with side dishes. 4️⃣ Why it's worth watching He notes a university lab livestreamed a training run days earlier, but at nowhere near this scale or frontier relevance. For anyone who wants to learn how production RL behaves — rewards, entropy, dynamic sampling, benchmarks moving step by step — this is a rare open classroom. His only complaint, half-joking: a livestream this good deserves a comment section. 🔗 Full Reading: #Xiaomi# #MiMo# #ReinforcementLearning# #PostTraining# #LLM# #OpenScience# #AI#

For anyone who is either a fan of vaccines, or a critic of vaccines but open-minded to the idea that vaccines *could* be great if they were made with a more open-source open-science and public-spirited approach, I highly recommend checking out PopVax.

Science can help build a more inclusive and sustainable future. A new @UNESCO exhibit at UNHQ in NYC shares the stories and aspirations of young researchers from across the globe, exploring open science, inclusion & gender equality.

New keynote alert! 🎤 Rachad Alao, VP of Engineering at @Cohere, joins #PyTorchCon# North America to explore how open science can help AI foundation models advance with transparency, security, speed, and scale. 📅 October 20-21 📍 San Jose Save your seat:

Freedom Tech DC: AI-Fueled Human Flourishing with @matthew_pines - AI could compress the path from scientific discovery to practical technology - Cheaper energy, new materials, and healthier lives are the biggest potential gains - Public support for AI infrastructure depends on sharing those gains - Open science means results anyone can check and tools anyone can build on (00:00) — Introduction (01:58) — Accelerating scientific discovery (03:08) — AI at the frontier of mathematics (04:11) — Energy, materials, and human flourishing (06:32) — The politics of AI infrastructure (07:17) — Automating discovery (07:59) — Open science and shared prosperity

71 nations have signed the Artemis Accords, committing to safe, transparent space exploration, and the timely release of scientific data. Read more about NASA's commitment to open science as we prepare to explore the Moon through @NASAArtemis:

Join me & NVIDIA’s @ctnzr (Bryan Catanzaro) Friday, Sept 11, 3 PM PT at the Open Source AI Summit in SF. Open science, model & artifact licensing, model performance, cyber risk, deployment architectures & agentic harnesses are all on the agenda. Apply:

🧬 Trending Now — SciPad is now live on Element🔥 Meet @scipadbsc — a launchpad for public science. 🔬 First experiment: Integer Factorization ⚡ Compute → Contribution 🧠 Open science → On-chain participation Explore SciPad on @Element_Market 👇 🔗 #SciPad# #NFT#