@jiqizhixin

–

16 April 2025 7h35

A Clean Slate for Offline RL
Paper: https://
arxiv.org/pdf/2504.11453

Oxford Unifloral Framework Advances Offline Reinforcement Learning

By

–

16 April 2025 7h35

Offline RL is a mess: unclear goals, tangled code, and sneaky online tuning A team from University of Oxford fixs it with: Unifloral — one clean framework, shared hyperparams
Result? New SOTA algorithms: TD3-AWR & MoBRAC.

ICLR Experiments with AI Agents for Meta-Reviewing Peer Feedback

By

–

16 April 2025 4h36

ICLR 2025 is experimenting with an AI agent that leverages multiple LLMs to provide optional feedback on reviewer reports. It’s an interesting move—using LLMs to meta-review human reviews. Future of peer review?

OpenAI Builds AI-Powered Social Platform to Compete with X

By

–

16 April 2025 2h40

According to The Verge, OpenAI is building its own version of 𝕏. A one-stop AI platform? Social + chat + tools? The lines are starting to blur…

Reasoning Models Effective Without Extended Thinking Process

By

–

16 April 2025 2h34

Reasoning Models Can Be Effective Without Thinking https://
arxiv.org/pdf/2504.09858

Reasoning models achieve top performance without extended thinking

By

–

16 April 2025 2h34

So apparently… Reasoning models can be effective without thinking! This flips the table on assumptions that long, detailed chains of thought are always necessary. Turns out, with the right training methods, even concise responses can deliver top-tier performance.

GPT-4.1 Prompting Guide: Strategies from OpenAI

By

–

15 April 2025 18h35

If you're looking for the best way to prompt GPT-4.1, you should definitely check out the GPT-4.1 Prompting Guide—straight from OpenAI, the folks who built it. It’s packed with practical strategies to get the most out of the model. https://
cookbook.openai.com/examples/gpt4-
1_prompting_guide
…

15 April 2025

AI Model Enhanced: Better Instruction-Following and Cinematic Aesthetics

By

–

15 April 2025 9h55

It now offers significantly better instruction-following, enhanced cinematic aesthetics, and richer artistic diversity. With support for over 60 stylized effect transfers, the model outputs are now more creative, imaginative, and visually compelling than ever. pic.twitter.com/0x7W6s4TEK
— 机器之心 JIQIZHIXIN (@jiqizhixin) 15 avril 2025

It now offers significantly better instruction-following, enhanced cinematic aesthetics, and richer artistic diversity. With support for over 60 stylized effect transfers, the model outputs are now more creative, imaginative, and visually compelling than ever.

15 April 2025

Kling 2.0 Launches with Enhanced Literary and Artistic AI Capabilities

By