RT @karpathy: Pleasure to come by the YC AI Startup School today! I'm told the recordings will be up "in the coming weeks", I'll link to it….

RT @JiaxunZhang6: ⚠️ Rogue AI scientists? 🛡️ SafeScientist rejects unsafe prompts for ethical discoveries. Check out paper ➡️ ( https://t….

4/ We're passionate about advancing time-aware AI! That's why we're releasing: .📚 Time-Bench: A large-scale (200k+ examples from 10 yrs of news) multi-task temporal reasoning dataset on 🤖 Time-R1 models: .#OpenSource.

3/ The results? Our 3B Time-R1 outperforms models over 200 times its size (like the 671B DeepSeek-R1) on highly challenging future event prediction & creative scenario generation tasks! This shows smaller, efficient models can achieve superior temporal performance.

2/ How do we do it? #TimeR1 uses a novel 3-stage RL curriculum, driven by a meticulously designed dynamic rule-based reward system. This progressively builds:.🧠 Foundational temporal understanding .🔮 Future event prediction skills .✨ Creative future scenario generation

1/ Why do LLMs often falter with time? 🤔 Key issues include: .➡️ The static nature of their training corpora, which inevitably become outdated. ➡️ Non-chronological training processes, where temporal info is handled out of sequence, hindering robust event-time logic.

RT @peixuanhakhan: (1/5) Want to make your LLM a skilled persuader?. Check out our latest paper: "ToMAP: Training Opponent-Aware LLM Persua….

💥Time-R1 is here! Can a 3B LLM truly grasp time? 🤔 YES! . Excited to share our new work, Time-R1: Towards Comprehensive Temporal Reasoning in LLMs 🚀. Check it out:.📖 Paper: 💻 Code: #TemporalReasoning #RL #LLMs

⚡️ Excited to share our PRX highlighted paper entitled "Early Predictor for the Onset of Critical Transitions in Networked Dynamical Systems"!.

RT @xinyazhang_: Excited to share our #NeurIPS2023 conference paper. 🔥🚀Attentive Transfer Entropy to Exploit Transient Emergence of Coupli….