Jun 17, 2025

• Unsupervised Elicitation of Language Models
  The Internal Coherence Maximization (ICM) algorithm fine-tunes pretrained language models using their own generated labels, achieving performance comparable to golden supervision and surpassing crowdsourced human supervision on various tasks. Link to article

• Nanonets-OCR-s
  Nanonets-OCR-s is a cutting-edge image-to-markdown OCR model that intelligently transforms documents into structured markdown, enhancing compatibility with Large Language Models (LLMs) for advanced processing.

• Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Task
  Cognitive debt accumulates when using AI tools like ChatGPT for essay writing, as evidenced by participants showing weaker neural connectivity and lower ownership of their work compared to those using traditional methods.

• We're expanding our Gemini 2.5 family of models
  Gemini 2.5 now includes Flash and Pro models, which are generally available, alongside the introduction of 2.5 Flash-Lite, the fastest and most cost-efficient model yet.

• The drawbridges come up: the dream of a interconnected context ecosystem is over
  MCPs are emerging as a pivotal technology, echoing the Web 2.0 vision of interconnected services, yet they face increasing restrictions that hinder competition and innovation.

• How you breathe is like a fingerprint that can identify you
  Breathing patterns are unique to individuals, akin to fingerprints, and can reveal insights into their physical and mental health, including correlations with BMI and anxiety levels.

• AMD's CDNA 4 Architecture Announcement
  AMD’s CDNA 4 architecture enhances matrix multiplication performance for machine learning by optimizing lower precision data types, while maintaining a competitive edge in vector operations through a familiar chiplet design.

• Programming Language Design in the Era of LLMs: A Return to Mediocrity?
  LLMs challenge traditional DSL design by raising the question of necessity; if LLMs can generate code efficiently, why invest in creating domain-specific languages that eliminate boilerplate?

• Vision Transformers Don't Need Trained Registers
  Vision Transformers can effectively operate without trained registers, as demonstrated in recent research that explores artifacts in attention and feature maps, paralleling findings in large language models.

• Breaking Quadratic Barriers: A Non-Attention LLM for Ultra-Long Context Horizons
  The proposed non-attention architecture for large language models (LLMs) can efficiently manage context windows of hundreds of thousands to millions of tokens, circumventing the quadratic memory issues of traditional Transformers.

• Time Series Forecasting with Graph Transformers
  Graph Transformers enhance time series forecasting by leveraging interconnected data from relational databases, allowing for more accurate predictions through a comprehensive end-to-end pipeline that integrates various signals from the entire graph structure.

• Research Scientists + Engineers for Generative AI at NVIDIA
  NVIDIA is seeking senior and principal research scientists to advance generative AI, focusing on training and deploying frontier-scale models and optimizing architectures for enhanced AI performance.

• Why Claude Code feels like magic?
  Claude Code feels like magic due to its iterative nature, allowing it to explore the entire solution space rapidly, enhancing user experience without increasing model intelligence.

• Voyager: Real-Time Splatting City-Scale 3D Gaussians on Your Phone
  Voyager enables real-time city-scale 3D Gaussian Splatting on mobile devices by streaming only the necessary Gaussians, significantly reducing data transfer needs and improving rendering speed.

• Real-time action chunking with large models
  Real-time chunking (RTC) enables robots to execute actions without delays or discontinuities, significantly improving performance in dynamic tasks like striking a match or plugging in an Ethernet cable, even under high latency conditions exceeding 300 milliseconds.