scale
PulseAugur coverage of scale — every cluster mentioning scale across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
Together AI outlines strategy for migrating to open-source models
Together AI's blog post outlines a strategy for migrating from closed-source to open-source AI models, emphasizing that such migrations can be faster and less complex than traditional ones, especially when utilizing man…
-
New AI model predicts cellular responses from unpaired data
Researchers have developed SCALE, a novel conditional transport model designed to predict cellular responses to perturbations. Unlike previous methods that required paired control and treated cell populations, SCALE can…
-
AfterQuery becomes YC's fastest unicorn at $3.2B valuation
AI training data startup AfterQuery has achieved unicorn status with a valuation of $3.2 billion, marking it as Y Combinator's fastest-ever to reach this milestone. This rapid ascent, occurring just five months after a …
-
Harvey Tenet: New Legal Agent Model Leverages Kimi K3 and Fireworks
Harvey has unveiled Harvey Tenet, a new research preview model specifically designed for long-horizon legal tasks. This model is built upon the Kimi K3 base and enhanced through asynchronous reinforcement learning using…
-
NVIDIA hires 109 Poolside AI staff in unique licensing deal
Poolside, a company known for its AI model factory, has undergone a significant shift involving NVIDIA and its founder Jensen Huang. Instead of a traditional acquisition, NVIDIA has licensed Poolside's technology and hi…
-
New SCALE method enhances AI planning by improving latent space geometry
Researchers have developed SCALE (State-Calibrated Latent Embeddings), a new method to improve planning in joint-embedding predictive world models. SCALE enhances the geometric properties of latent representations, simi…
-
New frameworks leverage LLMs and embeddings to expand scientific taxonomies · 2 sources tracked
Two new research papers introduce frameworks for enhancing scientific taxonomies using Large Language Models (LLMs) and embeddings. The first, ReLTEx, focuses on reliable LLM-based taxonomy expansion by combining LLM ge…
-
New AI framework tackles semiconductor design rule violations
Researchers have developed SCALE, a novel framework designed to address complex design rule violations in semiconductor manufacturing at advanced sub-2nm nodes. The system utilizes a self-supervised layout-generation st…
-
AI model quality shifts from scale to embedded reasoning
The era of AI model quality being solely dependent on scale is ending. Future advancements will focus on embedding reasoning directly into model weights, rather than relying on external tools like scratchpads or complex…
-
ConceptCF method enhances AI explainability for time series data
Researchers have introduced ConceptCF, a novel method for generating counterfactual explanations for time series data. This approach focuses on modifying human-interpretable concepts within the data, rather than individ…
-
New GAMUT benchmark tests AI factual completeness, Gemini 3.1 Pro leads
Researchers have developed GAMUT, a new benchmark designed to evaluate the factual completeness of long-form AI-generated text. Unlike previous methods that focused on the accuracy of individual claims, GAMUT assesses w…
-
VCs reveal 38-point framework for spotting unicorn founders
Venture capitalists Paige and Leura Craig have developed a 38-point framework to identify potential unicorn founders, focusing on traits like decision-making in chaos, resilience, and irrational optimism. Their approach…
-
Anthropic's Claude Fable 5 excels at complex tasks, scoring 16.1% on Remote Labor Index
Anthropic's Claude Fable 5 model has achieved a score of 16.1% on the Remote Labor Index, a benchmark designed to assess AI's capability in performing freelance computer work. This performance reportedly doubles that of…
-
New SCALE framework enhances coordination for heterogeneous robot fleets
A new framework called SCALE has been developed to address the challenge of coordinating heterogeneous robot fleets in industrial settings. This framework integrates real-time planning with robust execution by introduci…
-
New framework SCALE enhances robot fleet coordination in industrial environments
A new research paper introduces SCALE, a reactive online coordination framework designed to improve the efficiency of heterogeneous robot fleets in industrial settings. The framework addresses challenges in real-time pa…
-
CUDA emulator for AMD GPUs Zluda loses funding, reverts to hobby status
The Zluda project, an open-source effort to emulate NVIDIA's CUDA on AMD GPUs, has lost its commercial funding and reverted to a hobbyist status. Despite this setback, version 6 of Zluda introduces new features, includi…
-
Anthropic's Claude Code Lead: Recursive Agents Are Next Era of Coding
Boris Cherny, lead of Anthropic's Claude Code, discussed the future of coding at Meta's @ Scale conference. He highlighted recursive agent loops as the next significant development, describing them as continuously self-…
-
New benchmarks and models advance vision-language capabilities in robotics and reasoning · 10 sources tracked
Recent research explores advancements in vision-language models (VLMs) across several domains. DeCAL introduces a new model for dexterous manipulation that integrates tactile sensing and visual-language understanding. R…
-
New AI scheduler SCALE generalizes to unseen cluster sizes
Researchers have developed SCALE, a new deep reinforcement learning scheduler designed for agentic LLM systems that can manage tasks across heterogeneous clusters of varying sizes. Unlike previous schedulers that requir…
-
New SCALE method improves conformal prediction for graph-structured time series
Researchers have introduced a new method called Spectral Conformal prediction via wAveLEt transform (SCALE) to improve uncertainty quantification in forecasting graph-structured time series. Traditional conformal predic…