Name: [PDG 465] Kimi K2: Open Agentic Intelligence
Start: 2025-12-02T20:00:00+01:00
End: 2025-12-02T22:00:00+01:00

**Link to article**: https://arxiv.org/pdf/2507.20534
**Title:** Kimi K2: Open Agentic Intelligence
**Track:** RL
**Content**: Kimi K2 is a 1T-parameter Mixture-of-Experts LLM that uses the new MuonClip optimizer—an improvement over Muon that adds QK-clip—to enable stable, spike-free pretraining on 15.5T tokens with high token efficiency. After pretraining, the model undergoes multi-stage post-training, including a large agentic data-synthesis pipeline and a joint reinforcement learning (RL) stage where it learns from interactions with real and synthetic environments. K2 achieves state-of-the-art results among open-source non-thinking models, especially in coding, math, software engineering, and agentic tasks, and its base and post-trained checkpoints are publicly released.
**Slack link**: ml-ka.slack.com, channel: #pdg. Please join us -- if you cannot join, please message us here or to mlpaperdiscussiongroupka@gmail.com.

In the Paper Discussion Group (PDG) we discuss recent and fundamental papers in the area of machine learning on a weekly basis. If you are interested, **please** **read the paper beforehand** and join us for the discussion. If you have not fully understood the paper, you can still participate – everyone is welcome! You can join the discussion or simply listen in. The discussion is in German or English depending on the participants.

DavidFarago

AI Paper Discussion Group Karlsruhe

Technology

Data Science

Neural Networks

Deep Learning

Natural Language Processing

Artificial Intelligence

Machine Learning

Academics

Intellectual Discussions

Ahmed Anwari

CHRISTINA peters

Nidhi

Mario Zachmann

Lionel Chiron

Leonard Plotkin

r-stl

[PDG 465] Kimi K2: Open Agentic Intelligence

Online event

Teilen

AI Paper Discussion Group Karlsruhe

[PDG 465] Kimi K2: Open Agentic Intelligence

AI Paper Discussion Group Karlsruhe

Details

Mitglieder interessieren sich auch für