open-research

Will Brown 指出 Intellect-2 的训练配方（recipe）其实早在 2025 年 5 月就已发布，并附上了相关链接。

Banbury Road unveils Kardashev-0.7, a population of 32 distinct models trained jointly with reinforcement learning to pursue Population Scaling, treating model count and collaboration as a new scaling axis toward complementary specializations.

Meta AI researchers partnered with mathematicians to use Muse Spark 1.1/1.2 in Thinking Mode to tackle open research problems, releasing six papers — five of which present solutions to previously open mathematical questions. The collaborations followed strict attribution and human-review principles, with acknowledgment of independent concurrent solutions from other teams.

AI researchers Nathan Lambert and Tom Zick have founded Trillium Labs, a nonprofit aiming to conduct high-stakes AI research—including recursive self-improvement and agents—openly and transparently, countering the closed-development approach of frontier labs like OpenAI and Anthropic.

kyutai_labs released a technical blog on training PocketTTS, a 100M parameter on-device text-to-speech model using the drifting one-step generative objective from Deng et al., achieving under 1% WER with high quality and voice cloning — reportedly the first speech and autoregressive model trained this way.

The post questions why Chinese AI labs are achieving better results with less investment compared to American labs, considering factors like efficiency, post-training techniques, and data acquisition.

Dario Amodei highlights the AI alignment problem as a threat to humanity and urges AI firms to openly collaborate on safety research instead of treating it as a competitive advantage.

OpenWAM presents an open, modular framework for world-action model pretraining to systematically identify design principles, with pretrained models showing strong performance in simulation and real-world robotics.

The Bit Radix Theory argues that different forms of intelligence, such as human and AI, may have fundamentally distinct cognitive approaches rather than converging, emphasizing complementary strengths. The paper promotes open critical engagement by inviting readers to test the theory using AI tools.

PyTorchCon North America announces Joelle Pineau, Chief AI Officer at Cohere, as a keynote speaker. The event takes place October 20-21 in San Jose, CA, focusing on advancing AI through open research and collaboration.

MiniMax announces a planned open weights mini-march in San Francisco, promoting open AI models, research, and innovation.

Alphaxiv and Hugging Face launch a community challenge to test the reproducibility of AI research papers from ICML 2026, offering $4500 in GPU credits and an autoresearch agent to help participants.

A founder shares his experience turning down two acquisition offers (at $40M and $400M valuations) for an AI interpretability startup using a geometric and proprioceptive approach, now with a finished product and open research on Zenodo.

An open evaluation setup with 55 LLMs blind-grading each other reveals statistically significant same-family rating bias across 8 model families, with Mistral penalizing its own models most severely. The study highlights issues with aggregate leaderboards and proposes improvements like within-response mixed-effects models.