Absolute Zero: LLMs can train without any external data Has the "data wall" just been breached? Recent RL paradigms often relied on a set of questions an answers that needs to be manually curated. Researchers from @Tsinghua_Uni went like "why though". Indeed, why learn
RESEARCH
-

Transform Documents into High-Quality Synthetic Datasets with Llama
By
–
Turn any document into a high-quality synthetic dataset using Llama! Unsloth AI and Meta released a free notebook that transforms your documents into high-quality synthetic datasets, and then finetuning them using Llama. 100% Open Source
-
Meta’s Synthetic Data Kit for Llama 3.2 on Google Colab
By
–
Link to the Colab Notebook here: https://
colab.research.google.com/github/unsloth
ai/notebooks/blob/main/nb/Meta_Synthetic_Data_Llama3_2_(3B).ipynb
… Synthetic Data kit from Meta: -
ChatGPT vs Gemini vs Grok: AI Model Performance Comparison
By
–
by all counts, chatgpt is worst at reasoning while gemini or in that matter grok also outperforms in terms of latency, quality of output and research skills.
-

100 AI Scientists Establish Singapore Consensus Trustworthy AI
By
–
100 leading AI scientists map route to more 'trustworthy, reliable, secure' AI The landmark Singapore Consensus comes at a time when the giants of generative AI – such as OpenAI – are disclosing less and less to the public. https://
zdnet.com/article/100-le
ading-ai-scientists-map-route-to-more-trustworthy-reliable-secure-ai/
… @Yoshua_Bengio -
GPT-4.5: Ideal for Writing and Research
By
–
2. GPT-4.5 (Research Preview)
— God of Prompt (@godofprompt) 13 mai 2025
Great for exploring ideas and writing deeply.
Made for:
• Long-form writing
• Brainstorming
• Deep research & outlining
Use for:
→ Essays
→ Scripts
→ Strategy plans
→ Newsletter writing (mostly writing tasks) pic.twitter.com/fdNd1iYy3L2. GPT-4.5 (Research Preview) Great for exploring ideas and in-depth writing. Designed for:
• Long-form writing
• Brainstorming
• Deep research & outlining Use for:
→ Essays
→ Scripts
→ Strategy plans
→ Newsletter writing (primarily writing tasks) -

FaceAge: AI Tool Estimates Cancer Survival from Facial Photos
By
–
Mass General Brigham’s researchers introduced FaceAge, an AI tool that can estimate cancer survival outcomes with facial photos The AI estimates biological age from photos, helping teams guide their treatment levels accordingly
-

OpenAI Releases HealthBench for Evaluating AI in Healthcare
By
–
Et si vous aviez un médecin de classe mondiale… dans votre poche, 24h/24, gratuitement ? C’est la promesse de l’IA dans la santé. Mais ici, une seule erreur peut coûter une vie. C’est pour ça qu’OpenAI vient de publier HealthBench, un nouveau benchmark open-source qui évalue
-
RL Verification Rewards vs SFT Synthetic Data Training
By
–
this isn't right. RL with verification-based-rewards is a different learning signal than generating synthetic data via language modeling and then using SFT, which is what you're describing
-
From Supervised to Reinforcement Learning: The Next AI Frontier
By
–
the AI labs spent a few years quietly scaling up supervised learning, where the best-case outcome was obvious: an excellent simulator of human text now they are scaling up reinforcement learning, which is something fundamentally different. and no one knows what happens next
