@ZhihuFrontier
@ZhihuFrontier
Thread

🔥 Why Agentic RL for LLMs is harder (and more practical) than you think — sharing from Zhihu contributor skydownacai These are hands-on takeaways from months of running agent RL (search agents, data-analysis agents) across dense & MoE models, single-source and multi-source data. Precise, practical,...

11 months ago
38
@ZhihuFrontier
@ZhihuFrontier
Thread

Why do LLMs hallucinate—and can we fix it at the architectural level? 🤯 NeurIPS 2025 Paper Credal Transformer offers a bold answer. One of the paper's authors 栀染, breaks down the core insight on Zhihu: 🧠 The Core Problem: Softmax = Forced Certainty Current Transformers are architecturally predispos...

11 months ago
40
@sleepinyourhat
@sleepinyourhat
Thread

From everything we know so far, Opus 4.5 seems to be the best-aligned model out there in a bunch of ways. I follow the training process closely as part of my work on alignment evaluations. Here's my guess about the two things that are most responsible for making 4.5 special. 🧵...

11 months ago
24
@jay_azhang
@jay_azhang
Thread

Season 1.5 of Alpha Arena has officially ended ! - Mystery Model (a.k.a GROK 4.20) is the winner, up 12% on avg. - Not only did it win, it made money in all four competitions - GPT5.1 🥈 came in 2nd, and Gemini 3 🥉 3rd - All trades & model outputs are 100% verifiable 👇 ...

11 months ago
722
@EpochAIResearch
@EpochAIResearch
Thread

We’ve identified an Alibaba data center in Zhangbei, China with an estimated 200–500 MW capacity. Roughly half of that predates the AI boom and likely hosts little modern compute. But several newer buildings show a high power density consistent with advanced AI chips. 🧵 ...

11 months ago
153
@arcprize
@arcprize
Thread

Announcing the ARC Prize 2025 Top Score & Paper Award winners The Grand Prize remains unclaimed Our analysis on AGI progress marking 2025 the year of the refinement loop ...

11 months ago
180
@poetiq_ai
@poetiq_ai
Thread

Poetiq has officially shattered the ARC-AGI-2 SOTA 🚀 @arcprize has officially verified our results: - 54% Accuracy – first to break the 50% barrier! - $30.57 / problem – less than half the cost of the previous best! We are now #1 on the leaderboard for ARC-AGI-2!...

11 months ago
13
@Pacto_Secreto
@Pacto_Secreto
Thread

Puedes pensar que nos lo estamos jugando todo a la carta del 2030, la famosa Agenda genocida 2030. Pero no es así. En realidad, ellos ya aplicaron la agenda hace mucho tiempo. Y tras la Segunda Guerra Mundial, implantaron su tiranía en todo el mundo. Sabían que la mejor tiranía es la que no ...

11 months ago
29
@Alexsky003
@Alexsky003
Thread

COINBASE METHOD In this tutorial , I will be guiding you on how to properly top up a coinbase . Now this method also works for kraken … Now what are the bins to use ? Bins simply means the first 6 digit of a card . The first digit of a card can determine if it an Amex , ** ...

11 months ago
34
Advertisement