Aqua
PulseAugur coverage of Aqua — every cluster mentioning Aqua across labs, papers, and developer communities, ranked by signal.
- 2026-08-31 research_milestone Researchers developed AQuA, a system for recursively self-improving quantitative trading research agents that ensures reliable and reproducible results. source
2 day(s) with sentiment data
-
Princeton, Ant Group, Stanford unveil AQuA for autonomous finance research
Researchers from Princeton University, Ant Group, and Stanford University have developed AQuA, a novel two-part agentic framework designed to autonomously discover factors and develop models in quantitative finance. The…
-
New research reveals sparse and structured nature of effective LoRA writes
Researchers have found that effective Low-Rank Adaptation (LoRA) updates in language models are surprisingly sparse and structured, rather than uniformly distributed across parameters. Using a technique called Learned-B…
-
AI agents for quantitative research gain reliability with AQuA system
Researchers from Princeton University, Ant Group, and Stanford University have developed AQuA, a system designed to improve the reliability of AI agents in quantitative research. AQuA separates the research process into…
-
AQuA research clarifies 'self-improvement' does not rewrite agent LM weights
A discussion on Reddit clarifies that the AQuA research paper details a "recursive self-improvement" process for a bounded research loop, but it does not involve the agent's language model rewriting its own weights. The…
-
AI usage remains unclear; new research highlights strong results and failures
The Download newsletter from MIT Technology Review discusses how people are currently using AI, noting that the exact applications remain unclear. Separately, a paper on Aqua presents strong results, with the authors tr…
-
New Semi-CoT Framework Enhances LLM Reasoning with Pseudo-Supervision
Researchers have introduced Semi-CoT, a novel framework for Semi-supervised Chain-of-Thought Learning that leverages unlabeled questions to generate pseudo reasoning supervision. This method refines the self-training ap…
-
AI dictation apps leverage LLMs for improved accuracy and features
Several AI-powered dictation applications have emerged, offering significant improvements in accuracy and context-awareness due to advances in LLMs and speech-to-text models. These tools now automatically handle formatt…
-
Eleven Labs, Cohere, Grok lead in AI model quality benchmarks
A recent comparison of speech-to-text models highlights Eleven Labs' Scribe v2 as the top performer with a score of 20,251. Cohere's model followed closely at 19,885, with Grok achieving 19,611. AssemblyAI's Universal 3…