Can LLMs achieve massive capacity gains without massive increases in computational cost? YES, say researchers from Shanghai Jiao Tong University and Xiaohongshu! They introduce JTok, a novel scaling method that uses lightweight "token-indexed parameters" to intelligently
RESEARCH
-
Evaluation awareness in Opus 4.6: measurement validity concerns
By
–
Eval awareness in Opus 4.6 is a bit alarming TBH. If the model behaves differently when it thinks it's being tested, what are we actually measuring?
-
Micro-benchmarks Don’t Measure True Reasoning Capabilities
By
–
These micro-benchmarks are fun but I've found the model that "wins" changes depending on the exact framing of the prompt. Pattern matching != reasoning.
-
Lending stochastic parrots term, refining critique
By
–
Yeah, we’ve got to say it! (Well, after that, I did lend him the term “stochastic parrots,” which he doesn’t use, thanks @Fabien_Mikol
, I need to refine my critique) -
Grok’s Learning: Interview Shifts AI Development Timelines and Bottlenecks
By
–
Why do AI coding tools score high on tests, but don't always help developers work faster? This @DigEconLab talk presents the evidence on this productivity paradox, bottlenecks in deployment, and next steps for understanding AI’s productivity impacts:
-
Framework for Understanding Technical Organization Metrics
By
–
A framework for making sense of metrics in technical organizations https://
buff.ly/2KXCf7N
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
SigLIP 2: Google’s Advanced Vision-Language Model Evolution
By
–
SigLIP 2: Advancing Vision-Language Understanding Without Contrastive Bottlenecks
— Satya Mallick (@LearnOpenCV) 13 mars 2026
In this episode of Artificial Intelligence: Papers and Concepts, we explore SigLIP 2, the next evolution of Google’s vision–language model designed to better connect images and text through… pic.twitter.com/7TysLsvoXuSigLIP 2: Advancing Vision-Language Understanding Without Contrastive Bottlenecks In this episode of Artificial Intelligence: Papers and Concepts, we explore SigLIP 2, the next evolution of Google’s vision–language model designed to better connect images and text through
-
AI Panel Discussion on Economic Impact and Productivity Growth
By
–
As you can tell from the photo, we had some fun on this @SIEPR panel discussion about AI and the economy. And yes, we also discussed some serious topics, from productivity growth and economic disruption to catastrophic risk and the need for better metrics.
-
India Launches Param2-17B LLM Under BharatGen Initiative
By
–
Advancing India’s Sovereign AI Capabilities At the IndiaAI Impact Summit 2026, the Hon’ble Prime Minister had unveiled BharatGen Models of which comes the Param2-17B, a powerful large language model developed under the BharatGen initiative, marking a major step forward in
-
Free ML System Design Case Studies Repository 80+ Companies
By
–
A free repository of curated ML system design case studies from 80+ leading companies: https://
bit.ly/4kMcGbQ v/
@techNmak