Search agents could notice their own mistakes and fix their reasoning mid-trajectory. ReSeek enables exactly that. With a JUDGE action for on the fly self correction and dense rewards for factual correctness plus real utility, it trains agents that outperform SOTA on complex
MACHINE LEARNING
-
Self-Critique in Agentic AI: Benefits and Risks Explained
By
–
Self-critique is a core pattern in agentic AI — but it turns out to be dangerous in the wrong place.
New blog from @ArminPCM show: Self-critique helps when models are failing It destroys accuracy when models are already right
This has big implications for RLAIF, RLFT, and -
Self-Verification as Adversarial Training Signal for AI Models
By
–
The key takeaway:
Critique is for debugging — not polishing.
When the same model plays both “actor” and “judge,” self-verification can become an adversarial training signal.
Read the full breakdown → -
ScaleNet: Scaling Vision Transformers with Weight Sharing
By
–
What if you could scale up a ViT in minutes instead of training a giant one from scratch? ScaleNet makes it possible by inserting weight shared layers into pretrained ViTs with tiny adapter adjustments. It expands depth with almost no extra parameters, boosts a 2x depth DeiT
-
Apache Spark and Ray accelerate 1400 forecasting models training
By
–
Using Apache Spark™ for large-scale data prep and Ray for task parallelism, teams trained and tuned 1,400+ forecasting models in under 30 minutes instead of nearly three hours — hitting ~95% CPU utilization. See the full benchmark and workflow: https://
databricks.com/blog/superchar
ging-ai-model-building-data-and-task-parallelism-ray-and-databricks?utm_source=twitter&utm_medium=organic-social
… -
PyTorch vs JAX: Framework Comparison for Deep Learning
By
–
more of this energy for pytorch gives you jax. soo….
-
7 tips everyone ignores about ChatGPT
By
–
You think you know everything about #ChatGPT? Well no, here are 7 tips that everyone ignores → https://youtu.be/SJTQufPavHY #ChatGPT5 #GPT5 #GPT51
-
SAS and Quasar Transform Manufacturing with Predictive Analytics
By
–
SAS and Quasar are driving the future of AI in manufacturing. Predictive analytics is powering smarter, more efficient operations.
-
Machine Learning Community Unable to Handle Truth
By
–
Clearly the machine learning community can’t handle the truth. Good to see that @MrEwanMorrison can.
-
LLMs World Models Gap: Ongoing Critique and Arguments
By
–
among other places on facebook immediately after my deep learning is hitting a wall paper where i laid out the arguments in question, and also for years here on twitter, eg in 2019 on X calling my critique that LLMs lacked world models a “rear-guard action” see my recent