A big thank you to @itsmaddox_j for the invitation, and to professors @niclane7 and @contactrika for having me! slideshare.net/slideshow/int…
→ View original post on X — @maximelabonne, 2026-03-06 10:46 UTC
PROMPT ENGINEERING
-
SAGE-RL: Simple RL Training Modification
By
–
then they fold it into training with SAGE-RL. dead simple modification: in standard reinforcement learning (GRPO), you sample 8 responses per question. SAGE-RL replaces 2 of those 8 with SAGE-generated samples. the other 6 stay normal. one-line code change. the model learns to
-
RFCS Metric Reveals Early Correct Steps
By
–
first, the problem quantified. the researchers created a metric called RFCS (Ratio of First Correct Step) that tracks where in a chain of thought the correct answer first appears. on MATH-500, across every model tested, the right answer shows up well before the end in over half
-
WordPress Categories for AI News Aggregator
By
–
This is my first attempt at subscribers thread. 1/2
-
Troubleshooting AI Image Generation Filters and Prompt Issues
By
–
New @ScienceMagazine Engineering the brain's astrocytes to clear amyloid plaque! "CAR-A," in mouse model http://
science.org/doi/10.1126/sc
ience.ads3972
… http://
science.org/doi/10.1126/sc
ience.aef8451
… -
Using AI for Writing Tweets and Getting Blocked
By
–
Check out @PreciselyData — a leader in data integrity and Agentic-Ready Data for AI …
-
GPT-5.4 improves coding, search, and interruptions
By
–
BREAKING 🚨: GPT-5.4 delivers a major improvement across coding, web search, and many other areas.
— 🚨 AI News | TestingCatalog (@testingcatalog) 5 mars 2026
– SWE-Bench Pro – 57.7%
– OSWorld-Verified – 75.0%
– BrowseComp – 82.7%
Users can now interrupt the model and add instructions or adjust its direction mid-response.
GPT-5.4 is… https://t.co/vJNRakd1bN pic.twitter.com/Ap6jwBqcM7BREAKING : GPT-5.4 delivers a major improvement across coding, web search, and many other areas. – SWE-Bench Pro – 57.7%
– OSWorld-Verified – 75.0%
– BrowseComp – 82.7% Users can now interrupt the model and add instructions or adjust its direction mid-response. GPT-5.4 is -
GPT-5.4 Enhanced: Faster, More Factual AI with Thinking Interrupts
By
–
GPT-5.4 is our most factual and efficient model: fewer tokens, faster speed. In ChatGPT, GPT-5.4 Thinking has improved deep web research, better context retention when it thinks for longer—and oh—you can now interrupt the model and add instructions or adjust its direction
-
Man Builds Camera Hat, Grok Explains It to a Nine-Year-Old
By
–
This guy built a hat with cameras.
— Robert Scoble (@Scobleizer) 5 mars 2026
What does it all mean? I asked Grok to explain it like if I was a nine-year-old: https://t.co/yjJWALOLqV
Helping translate nerd to English. 🙂 https://t.co/dXH0QEditSThis guy built a hat with cameras. What does it all mean? I asked Grok to explain it like if I was a nine-year-old: https://
x.com/i/grok/share/0
7288a1006aa4d20958b173fd3a64315
… Helping translate nerd to English. 🙂 -
Utopai Studios unveils PAI for long-form video generation
By
–
BREAKING 🚨: Utopai Studios announces PAI, a new model for long-form video generation.
— 🚨 AI News | TestingCatalog (@testingcatalog) 5 mars 2026
Utopai offers a new workflow that expands a simple prompt into a comprehensive video scenario with detailed characters and keyframes.
Example 55-second movie, one-shot generation 👀 pic.twitter.com/7LyQOe5yFlBREAKING : Utopai Studios announces PAI, a new model for long-form video generation. Utopai offers a new workflow that expands a simple prompt into a comprehensive video scenario with detailed characters and keyframes. Example 55-second movie, one-shot generation