RT @DKokotajlo: I'm very happy to see this happen. I think that we're in a vastly better position to solve the alignment problem if we can….

RT @jmbollenbacher: The AI world needs to learn this lesson before its too late. We don't want an Oppenheimer amongst us. Physics has ne….

RT @davisblalock: While I briefly have no employer, let me tell you what's really happening with AI companies training on public data: [1/n].

RT @dlwh: So about a month ago, Percy posted a version of this plot of our Marin 32B pretraining run. We got a lot of feedback, both public….

Intuitively Gemma approach makes more sense, keeping the residual stream truly residual.

RT @kyutai_labs: Kyutai Speech-To-Text is now open-source! It’s streaming, supports batched inference, and runs blazingly fast: perfect for….

RT @g_k_swamy: It was a dream come true to teach the course I wish existed at the start of my PhD. We built up the algorithmic foundations….

I feel like OpenAI voice mode changed the tone of voice (not the persona) to have a different laid back cool vibe overnight. Don’t know if that is actually the case but if so it not good to do so without giving some kind of notice.

flops per capita.

RT @adibvafa: Introducing the world's first reasoning model in biology!. 🧬 BioReason enables AI to reason about genomics like a biology exp….

anthropic's "race to the top" is not going too well @AnthropicAI

I haven't been so worried about videogen effect on misinformation because there are relatively straightforward technological solutions where public figures can verify clips etc. But still, it is now in 2025 that they have to be rolled out.

RT @lifan__yuan: We always want to scale up RL, yet simply training longer doesn't necessarily push the limits - exploration gets impeded b….

RT @shaneguML: #veo3 is truly incredible. Here's my old explanation on why video/audio progress faster than text. 2025 is the year of agent….

RT @AdamRodmanMD: Huge update to our preprint today on the superhuman performance of reasoning models in medical diagnosis!. TL;DR – they d….

RT @MatternJustus: We went from our first line of code for prime-rl to releasing INTELLECT-2 in around two months. Now that our infra is in….

RT @ben_burtenshaw: Qwen3 Finetuning Notebook. I’m tuning @Alibaba_Qwen 3 for a fast local coding, and here’s a notebook for the process.….

RT @michael_nielsen: The implied point of view - "we are helpless passive entities unable to change this, merely to predict it" - kinda ann….

Interestingly they find that the 7B model learns some special things about cardinals that the large model does not and this makes it able to solve problems on PutnamBench that the big one does not.

They also distill the ProverV1.5-7B model on rollouts from this resulting large V3-derived model, and do further RL directly on it, calling it ProverV2-7B.