One of the most intriguing findings from the paper's analysis is that LLMs can autonomously devise random number extraction algorithms akin to hash functions (such as Sum-Mod or rolling hashes) within the context. The longer the inference model "thinks," the greater the
@sakanaailabs
-
Novel LLM Experts Discovery Through Task-Capability Coevolution
By
–
Discovering Novel LLM Experts via Task-Capability Coevolution
— Sakana AI (@SakanaAILabs) 21 avril 2026
Project: https://t.co/DT3tNbV5jz
Paper: https://t.co/3Nlz8ME43j
Can we build AI that is smarter than its parts? This week, our team will present AC/DC⚡ at #ICLR2026.
The current paradigm in AI assumes that to solve… pic.twitter.com/RBtnYjuvCFDiscovering Novel LLM Experts via Task-Capability Coevolution Project: https://
acdc-llm.github.io
Paper: https://
arxiv.org/abs/2604.14969 Can we build AI that is smarter than its parts? This week, our team will present AC/DC at #ICLR2026. The current paradigm in AI assumes that to solve -
String Seed of Thought: Prompting LLMs for Distribution-Faithful Generation
By
–
String Seed of Thought: Prompting LLMs for Distribution-Faithful and Diverse Generation Paper: https://
arxiv.org/abs/2510.21150
Blog: https://
pub.sakana.ai/ssot/ #ICLR2026 -
Can LLMs Generate Truly Random Distributions?
By
–
LLMは頭の中でコイントスができるか？
— Sakana AI (@SakanaAILabs) 20 avril 2026
ブログ：https://t.co/DNzwlrIJhQ
論文（#ICLR2026）：https://t.co/ZzJf6Q8FUb
一見簡単そうで奥深いこの問題を「プロンプトだけ」で解決した論文 "SSoT: Prompting LLMs for Distribution-Faithful and Diverse Generation" が #ICLR2026 に採択されました。… https://t.co/cLkasBwOc6Can LLMs Flip a Coin in Their Heads? Blog: https://
pub.sakana.ai/ssot Paper (#ICLR2026): https://
arxiv.org/abs/2510.21150 This seemingly simple yet profoundly deep problem was addressed in the paper "SSoT: Prompting LLMs for Distribution-Faithful and Diverse Generation," which has been -
Can LLMs Generate Truly Random Outputs Faithfully
By
–
Can LLMs flip coins in their heads?
— Sakana AI (@SakanaAILabs) 20 avril 2026
When prompted to “Flip a fair coin” 100 times, the heads to tails ratio drifts far from 50:50. LLMs can understand what the target probability should be, but generating outputs that faithfully follow a given distribution is a separate problem.… pic.twitter.com/XyF7Xnj8LlCan LLMs flip coins in their heads? When prompted to “Flip a fair coin” 100 times, the heads to tails ratio drifts far from 50:50. LLMs can understand what the target probability should be, but generating outputs that faithfully follow a given distribution is a separate problem.
-
EDINET-Bench: LLMs on Japanese Financial Statements
By
–
EDINET-Bench: Evaluating LLMs on Complex Financial Tasks using Japanese Financial Statements https://
arxiv.org/abs/2506.08762 https://
openreview.net/forum?id=Dxns0
cj15A
… #ICLR2026 -
Sakana AI’s EDINET-Bench Accepted to ICLR2026
By
–
Sakana AI's Japanese financial benchmark "EDINET-Bench," released last year, has been accepted to the international conference #ICLR2026. Blog: https://
sakana.ai/edinet-bench/ EDINET-Bench is a benchmark for evaluating LLMs across three tasks—accounting fraud detection, earnings -
Digital Ecosystems: Interactive Multi-Agent Neural Cellular Automata
By
–
Digital Ecosystems: Interactive Multi-Agent Neural Cellular Automatahttps://t.co/JgvytIgHH1 pic.twitter.com/HfqskCwcjF
— Sakana AI (@SakanaAILabs) 19 avril 2026Digital Ecosystems: Interactive Multi-Agent Neural Cellular Automata https://
pub.sakana.ai/digital-ecosys
tem/
… -
Competing Neural Networks Evolve in Digital Petri Dish Ecosystem
By
–
What happens when you put competing neural networks in a Petri Dish and start changing the rules while they adapt?
— Sakana AI (@SakanaAILabs) 18 avril 2026
Last year we released Petri Dish NCA, where neural nets are the organisms that learn during simulation. Today we're releasing Digital Ecosystems: a browser-based… pic.twitter.com/9UeAQ2mX1dWhat happens when you put competing neural networks in a Petri Dish and start changing the rules while they adapt? Last year we released Petri Dish NCA, where neural nets are the organisms that learn during simulation. Today we're releasing Digital Ecosystems: a browser-based
-
Sakana AI Featured for Defense AI Development Initiatives
By
–
We were featured on the Nikkei Podcast for our initiatives in the defense sector at Sakana AI. Under the theme "Entrusting Defense AI to Domestic Development: Japan's Sakana Secures Order for Analysis System," it is explained by Toyonori Nakanishi, Editor-in-Chief of NIKKEI