Solvita Framework Introduces Continuous Learning for LLMs in Competitive Programming
New agentic evolution framework enables LLMs to learn from previous problem-solving experiences without weight updates.
Sources · arxiv.org
The news desk
Follow what’s changing.
Recent developments in artificial intelligence, with the original sources one click away.
All news
801–820 of 1,902 Researchers Propose More Efficient Reinforcement Learning Methods for Post-Training Large Language Models
New research presents methods to improve RL efficiency for LLMs, reducing compute costs while maintaining performance across reasoning and coding tasks.
Sources · arxiv.org
Solvita Framework Enables Continuous Learning for LLMs in Competitive Programming Without Weight Updates
New agentic evolution framework allows LLMs to learn from past problem-solving experience using graph-structured knowledge networks.
Sources · arxiv.org
Three New Agent Frameworks Advance Domain-Specific AI Automation
Researchers publish frameworks for MAPDL simulation, colloidal packing, and self-evolving agent memory, addressing reliability and workflow automation.
Sources · arxiv.org
Three New ArXiv Papers Advance LLM Training and Recommendation Systems
Researchers propose methods to enhance cross-domain recommendations, optimize LLM systems with user logs, and improve reasoning through multi-stage training.
Sources · arxiv.org
Trust in OpenAI CEO Sam Altman Emerges as Central Theme in Musk-OpenAI Trial
The Elon Musk-OpenAI trial's final days focused heavily on questions about OpenAI CEO Sam Altman's trustworthiness, according to TechCrunch AI.
Sources · TechCrunch AI
ArXiv to Impose Year-Long Bans on Authors Who Rely Entirely on AI for Papers
Research repository ArXiv announces stricter enforcement against papers written entirely by large language models, introducing year-long author bans.
Sources · TechCrunch AI
Asexual Individuals Using AI Chatbots for Non-Sexual Intimacy, Report Finds
WIRED reports some asexual people are turning to AI companions for emotional connection without physical intimacy, though advocacy groups express concerns.
Sources · WIRED AI
OpenAI Co-Founder Greg Brockman Takes Charge of Product Strategy
Greg Brockman assumes product strategy role as OpenAI plans to integrate ChatGPT and Codex programming product.
Sources · TechCrunch AI
Greg Brockman Takes Control of OpenAI Products in Latest Executive Reorganization
OpenAI reshuffles leadership as Greg Brockman assumes control of products, aiming to unify ChatGPT and Codex into single core experience.
Sources · WIRED AI
LEMON Framework Advances Multi-Agent AI Orchestration Using Counterfactual Reinforcement Learning
New research introduces LEMON, an LLM-based system that optimizes multi-agent coordination through counterfactual learning signals.
Sources · arxiv.org
Musk v. Altman Trial Enters Final Week with Credibility Challenges
The third week of the Musk v. Altman trial saw lawyers attacking both parties' credibility as the case nears jury deliberation.
Sources · MIT Technology Review
New AI Research Advances Test-Time Adaptation, Search Skills, and Cross-Modal Generation
Recent arXiv papers introduce query-conditioned self-training, evolving skill banks for search, and zero-pair video-to-music generation methods.
Sources · arxiv.org
Researchers Advance Multi-Task Learning and Reasoning in Large Language Models
New research demonstrates significant improvements in multi-task learning efficiency and dialogue consistency through novel prompting and memory techniques.
Sources · arxiv.org
AI Safety Controls Remain Easily Bypassed Three Years After ChatGPT Launch
According to the New York Times, manipulating AI systems to produce harmful content remains trivially easy despite three years of safety development.
Sources · NYT Technology
Financial Services Face Data Readiness Challenge for Agentic AI Adoption
MIT Technology Review highlights how data preparation, rather than AI sophistication, is key to agentic AI success in highly regulated financial services.
Sources · MIT Technology Review
New Research Advances Multi-Agent AI Systems with Cognitive Models and Communication Frameworks
Four new papers explore improving LLM-based multi-agent systems through cognitive grounding, mechanism evaluation, world model alignment, and context learning.
Sources · arxiv.org
OpenAI Enables Codex Access Through ChatGPT Mobile App
OpenAI announces Codex integration with ChatGPT mobile app, allowing users to monitor and manage coding tasks remotely across devices.
Sources · OpenAI
Researchers Advance Test-Time Adaptation Methods for Large Language Models
New frameworks enable AI models to adapt during inference through query-specific training and multimodal reasoning without requiring additional training data.
Sources · arxiv.org
Google AI Reportedly Surfacing Users' Personal Phone Numbers Without Easy Opt-Out
A Reddit user reported receiving numerous unwanted calls after Google AI allegedly surfaced his personal phone number, with no clear way to prevent it.
Sources · MIT Technology Review