hardmaru
@hardmaru
Followers
347K
Following
144K
Media
4K
Statuses
26K
Building Collective Intelligence @SakanaAILabs 🧠
Minato-ku, Tokyo
Joined November 2014
The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery 👩🔬. It’s common for AI researchers to joke amongst themselves that “now all we need to do is figure out how to make AI write the papers for us!” but I think we’re now getting there!.
Introducing The AI Scientist: The world’s first AI system for automating scientific research and open-ended discovery!. From ideation, writing code, running experiments and summarizing results, to writing entire papers and conducting peer-review, The AI
33
145
760
RT @adcock_brett: Sakana AI dropped AB-MCTS, an algo that lets competing AI models work together, building on their strengths and errors, t….
0
17
0
Nice video discussing the recent thought-provoking paper: “Questioning Representational Optimism in Deep Learning: The Fractured Entangled Representation Hypothesis” By @akarshkumar0101 @jeffclune @joelbot3000 @kenneth0stanley. 15min @MLStreetTalk video↓.
AI is so smart, why are its internals 'spaghetti'? We spoke with @kenneth0stanley and @akarshkumar0101 (MIT) about their new paper: Questioning Representational Optimism in Deep Learning: The Fractured Entangled Representation Hypothesis. Co-authors: @jeffclune @joelbot3000
10
19
171
Someone like Soham Parekh would never be able to pass Sakana AI’s hardcore job application process. 😎🔥.
2
0
23
The top open SWE models are all Qwen3, Qwen2, QwQ, Deepseek-V3 or R1-based.
1
0
12
DeepSWE is a new state-of-the-art open-source software engineering model trained entirely using reinforcement learning, based on Qwen3-32B. Fantastic work from @togethercompute @Agentica_‼
Announcing DeepSWE 🤖: our fully open-sourced, SOTA software engineering agent trained purely with RL on top of Qwen3-32B. DeepSWE achieves 59% on SWEBench-Verified with test-time scaling (and 42.2% Pass@1), topping the SWEBench leaderboard for open-weight models. Built in
11
42
256
RT @SakanaAILabs: Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search. https://….
0
11
0
Our paper: “Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search”.
2
3
24
RT @MLStreetTalk: Very interesting work by @SakanaAILabs - they have designed a MoE / novel test time inference framework inspired by MCTS….
0
26
0
The Multi-LLM AB-MCTS combination of o4-mini + Gemini-2.5-Pro + DeepSeek-R1-0528, current frontier AI models, achieves strong performance on the ARC-AGI-2 benchmark, outperforming individual models by a large margin. Implementation of AB-MCTS on GitHub:.
3
7
54
Many ARC-AGI-2 examples that were unsolvable by any single LLM were solved by combining multiple LLMs. In some cases, an initially incorrect attempt by o4-mini is used by R1-0528 and Gemini-2.5-Pro as a hint to get to the correct solution. ARC-AGI-2 code:.
2
7
66
RT @SakanaAILabs: We’re excited to introduce AB-MCTS!. Our new inference-time scaling algorithm enables collective intelligence for AI by a….
0
217
0
Inference-Time Scaling and Collective Intelligence for Frontier AI. We developed AB-MCTS, a new inference-time scaling algorithm that enables multiple frontier AI models to cooperate, achieving promising initial results on the ARC-AGI-2 benchmark.
17
94
537