Back to people
@iwiwi
T

Takuya Akiba

音声
@iwiwi

Research Scientist @SakanaAILabs

25KFollowers1.5KFollowing1.2KPostsView on X

Recent posts

チームのLLM開発が本格化してきたので仲間を募集します! 人数を必要以上に増やさず、一人ひとりが大きな役割を担うチームを作っています。高い基礎力と知的好奇心、そして違った強みを持つメンバーが集まって、全体の目的に向かって高い主体性を持って動いています。そんなチームの一員になれます。 僕自身もこのチームでメンバーと仕事することが何より楽しいです。 LLM経験者はもちろん歓迎。未経験でも、しばらくLLMに人生をかけてみたいという気持ち、他分野での強みや実績、新しい領域を速く学べる基礎力がある人は是非と思ってます。

@SakanaAILabs
S
Sakana AI@SakanaAILabs

【採用情報】Member of Technical Staff (LLM Development) の募集を開始しました🐟 https://t.co/XmUoUaR3QY Sakana Namazuを起点に拡大しているSakana AIのLLM開発を担うポジションです。下記のような挑戦が待っています。 ・事前学習から事後学習・評価まで、LLM開発の全工程に取り組むチーム に参加 ・LLM開発全体を見渡しながら、担当領域にオーナーシップを持ってリード ・Sakana AIのプロダクト・ソリューションとRSI研究の基盤となるLLMを開発 LLM開発の最前線で挑戦したい方、ぜひご応募ください🚀

Photo 1

Sakana NamazuがOpenRouterに登場 https://openrouter.ai/sakana/sakana-namazu

@OpenRouter
O
OpenRouter@OpenRouter

Sakana Namazu by @SakanaAILabs is live on OpenRouter. Built on Kimi K2.6, Sakana Namazu is a specialized model that combines a deep understanding of Japanese culture with high-performance reasoning capabilities, integrating web search and code execution to handle complex tasks within a business context. https://t.co/YlHQNKtx3J

オフィスで一番着てる人の多いTシャツは何故か確実にModalのTシャツ お世話になっております 🙏

@modal
M
Modal@modal

Congrats to Sakana AI on shipping Namazu! Happy to power Namazu's ~1T-param model for live web search + code execution on Modal.

帰宅。 #ICFPC2026 お疲れ様でした。個人的にはKaggleのNeuroGolf 2026やNeurIPS 2025 Code Golfに通じるところを感じて、それらを参考にしつつCodex + {GPT5.6 Sol, Fugu-Ultra v1.1} をあれこれ工夫しながら叩いてました。結果は振るわず反省も多いが、とりあえず楽しかった!運営の方々には感謝🙏

@imos
い
いもす@imos

今年もICFPC2026のチームUnagi(@iwiwi, @sulume, @wata_orz, @toslunar, @chokudai)で参加しました。リポジトリとビジュアライザも公開しておきました! https://github.com/icfpc-unagi/icfpc2026 https://icfpc-unagi.github.io/icfpc2026/

72時間のコンテスト折り返し、チームUnagi現在3位

Photo 1
@iwiwi
T
Takuya Akiba@iwiwi

出ます!

新刊『検索システム』を著者の佐藤竜馬先生 @joisino_ よりご恵贈頂きました!検索の基本からベクトル検索、RAG、LLMによる生成検索まで幅広く扱っていて良さそうです。ありがとうございます!🙏 https://www.amazon.co.jp/dp/4065429714

Photo 1Photo 2

#ICML2026 でと発表したUnMaskForkについてのブログを出しました。複数の拡散言語モデルを協調させ推論時スケーリングする手法です。 @takkyuO2 との共同研究です。 この論文が出来るまでの過程はなかなか面白かったです。まず興味深い発見として、DreamCoder等のMDLMでは定番の「温度による多様性」 がほぼ使えませんでした。温度を0よりぐっと上げたり、デコード手法を触って確率性を入れようとすると、品質が急落するんですよね。そこで、複数の拡散言語モデルを混ぜ合わせるという少し奇妙な方法を試したところ、これが驚くほどうまく機能したという。 以前発表したAB-MCTS (NeurIPS'25)に続く論文となり「複数モデルの協調+推論時スケーリング」シリーズが作れたのも嬉しいです。

@SakanaAILabs
S
Sakana AI@SakanaAILabs

Can test-time scaling work for diffusion language models? In our #ICML2026 paper "UnMaskFork," we show that having multiple masked diffusion language models collaborate on a single answer improves performance on coding and math tasks. Blog: https://t.co/FZ25e6XCws Test-time scaling is an actively researched technique that boosts LLM performance by using inference-time compute, for example, by having a model think longer or repeatedly refine its answers. This allows us to enhance performance simply by increasing computation during inference without relying on additional training, giving us the flexibility to balance compute costs and performance based on the specific use case. Unlike standard LLMs that generate text left-to-right, masked diffusion language models (MDLMs) generate text by gradually filling in a fully masked sequence. MDLMs can generate multiple parts of a sequence in parallel, offering potential speed-ups, and they can generate flexibly while seeing the entire sequence at once. This makes them an actively studied new paradigm in language modeling. We found that the standard LLM approach of "raising the temperature to increase randomness and generate diverse answers" does not work well for MDLMs like Dream-Coder. Instead of relying on this randomness, our proposed method, UnMaskFork (UMF), creates diversity through "model switching." Multiple MDLMs share the task of unmasking a single answer, and we use Monte Carlo Tree Search to search for a promising sequence in which different models handle different stages. Each model picks up where the others left off, filling in the parts it is most confident about. This collaborative approach allows us to explore diverse answers while maintaining generation quality, consistently outperforming existing test-time scaling methods on coding benchmarks and scaling effectively on math as well. Test-time scaling is also crucial for advancing MDLMs, and our work shows that UMF can sidestep the difficulties specific to them. UMF requires no additional training or changes to the models; it works simply by combining pre-trained models at inference time. This allows us to leverage the diversity of diffusion language models trained on different data and with different methods to improve performance. We believe the value of UMF will only grow as more diverse MDLMs emerge. This work is part of our broader research into "collective intelligence of AI," alongside methods like AB-MCTS and Sakana Fugu that have multiple LLMs collaborate. We'll continue pursuing research that turns model diversity into a source of strength. For details of the algorithm and illustrative examples showing how this collaboration works, please see our blog and paper. Paper: https://t.co/4JC9SYdTyX 🐟

We’re at Hall A #1701 right now! Come discuss how we ensemble diffusion LMs for test-time scaling. #ICML2026

@SakanaAILabs
S
Sakana AI@SakanaAILabs

"UnMaskFork: Test-Time Scaling for Masked Diffusion via Deterministic Action Branching" will be presented at #ICML2026 Paper: https://t.co/4JC9SYdTyX We introduce UnMaskFork, a test-time scaling framework for Masked Diffusion Language Models (MDLMs). Using Monte Carlo Tree Search, it explores diverse generation paths by dynamically switching between multiple pre-trained MDLMs to collaboratively generate the text. Evaluations on coding and mathematical reasoning tasks show that UnMaskFork consistently outperforms standard Best-of-N and other tree search baselines. The results demonstrate that deriving search space diversity from multiple distinct models is a highly effective test-time scaling strategy for MDLMs.