T2R-bench A Benchmark for Generating Article-Level Reports from Real World Industrial Tables
MACHINE LEARNING
-
PVPO: Pre-Estimated Value-Based Policy Optimization for Agentic Reasoning
By
–
PVPO Pre-Estimated Value-Based Policy Optimization for Agentic Reasoning
-
Brain as Recurrent Neural Network: Understanding Biological Learning Algorithms
By
–
Fundamentally, the brain is a recurrent neural network. Maybe context windows and “notes to self” are sufficient, but it would be good to understand the learning algorithms that make biological recurrent networks as effective as they are, and we clearly don’t yet.
-
AI Progress on IMO Gold Medal Exceeded 2022 Expectations
By
–
We are way ahead of the curce. Only a minority of AI experts had expected in 2022 that gold would be achieved in the IMO as early as 2025; they were hugely mistaken, as it came faster than expected.
-
AI Math Performance Surpasses 2022 Expert Forecasts
By
–
We can now say pretty definitively that AI progress is well ahead of expectations from a few years ago. In 2022, the Forecasting Research Institute had super forecasters & experts to predict AI progress. They gave a 2.3% & 8.6% probability of an AI Math Olympiad gold by 2025…
-
High-Resolution Image Models Enable Advanced OCR and Visual Understanding
By
–
These models handle high-res images with ease and can be prompted for tasks like OCR, captioning, visual Q&A, and broader image understanding.
-
AI-powered prospecting identifies ICPs, enriches profiles, prioritizes leads
By
–
Step 1: AI-powered prospecting Instead of scraping random lists, agencies are using AI to: • Identify ICPs across multiple platforms
• Enrich profiles with context (budget, pain points, tech stack)
• Prioritize leads that actually buy No more spray-and-pray. -
Apple Launches FastVLM and MobileCLIP2 Open Models
By
–
Apple launched two new open models: FastVLM and MobileCLIP2
— The Rundown AI (@TheRundownAI) 2 septembre 2025
The models are up to 85x faster and 3.4x smaller than previous work
Suitable for high-res image processing tasks like OCR, image captioning, visual question answering, and image understandingpic.twitter.com/uG9ObsWACuApple launched two new open models: FastVLM and MobileCLIP2 The models are up to 85x faster and 3.4x smaller than previous work Suitable for high-res image processing tasks like OCR, image captioning, visual question answering, and image understanding
-
Google DeepMind August Releases: Gemini, Veo, Imagen and More
By
–
The craziest Google I/O to date! In August at Google DeepMind, it was: Nano Banana (Gemini 2.5 Flash Image)
Gemini Embedding
Veo 3 Fast
Genie 3
Imagen 4 Fast
Gemma 3 270M
Perch 2
Kaggle Game Arena
Gemini API Url Context
AI Studio Builder (UI overhaul, prompt suggestions, GitHub -
AGI Timeline Debate: Continuous AI Advancement Beyond 2030
By
–
The realistic alternative to "AGI by 2030" is not "AI goes away" but rather continued advancement in an industry with large-scale economic value (at least by early revenue numbers) and many smart people focused on figuring out new approaches that might work if old ones fail.