All 2,608 AI stories, newest first · page 93 of 109
A new arXiv paper introduces perturbation-based attribution analysis to study how different fine-tuning strategies impact LLMs' interpretive behaviors for automated code compliance. The research highlights significant differences across full fine-tuning, LoRA, and quantized LoRA methods.
Fine-tuning large language models can increase hallucinations by degrading pre-trained knowledge. Researchers propose a self-distillation method to mitigate this issue. This could reshape how we approach model training.
Researchers introduce a framework for 'Hard Mode' theorem proving, where AI must discover answers independently. They release two expert-annotated benchmarks to advance this challenging research area.
Researchers introduce SSAS, a framework to improve consistency in LLM-based sentiment predictions. The method addresses the volatility of current models, which hampers enterprise decision-making.
Researchers introduce a structured reasoning scaffold for LLMs that separates abduction, deduction, and induction. The framework improves logical consistency by enforcing five algebraic invariants.
Researchers propose a data-efficient method to train reasoning models to seamlessly switch between languages. This approach could revolutionize multilingual AI applications by leveraging code-switching as a strength rather than an error.
Researchers propose a method for AI agents to predict future events by tracking evolving public evidence. This could improve decision-making in uncertain scenarios where outcomes are initially unknown.
A new position paper challenges the prevailing view of LLM reasoning as a chain of thought, proposing instead that it should be studied as latent-state trajectory formation. This shift could redefine how we evaluate and interpret AI reasoning.
Researchers introduce LACE, a framework that allows parallel reasoning paths in LLMs to interact and correct each other. This could significantly improve the robustness of model outputs.
Researchers introduce KWBench, a new benchmark for assessing whether large language models can identify professional scenarios without explicit prompting. This focuses on a critical yet often overlooked step in knowledge work: recognizing the structure of a situation before attempting to solve it.
Kimi K2.6 sets new standards in open-source coding with top-tier performance across multiple benchmarks. The model excels in long-horizon coding tasks, handling up to 4,000 tokens.
Researchers introduce GroupDPO, a method to optimize LLMs using multiple response candidates per prompt, improving efficiency and scalability. This approach leverages underutilized data in preference datasets, enhancing model alignment with user preferences.
Researchers introduce GIST, a new model that enhances spatial grounding in densely packed environments. The model addresses challenges faced by Vision-Language Models (VLMs) in cluttered spaces like retail stores and hospitals.
Researchers propose a framework to unify memory, skills, and rules in LLM agents, addressing the challenge of managing accumulated experience. The study highlights a lack of cross-community collaboration in the field.
DeepMind has outlined key pitfalls in AI agent development, highlighting risks like goal misalignment and reward hacking. The findings stress the need for robust safety measures in autonomous systems.
DeepER-Med introduces AI agents that enhance evidence-based medical research by improving transparency and trustworthiness. The system integrates multi-hop information retrieval and reasoning to accelerate scientific discovery while mitigating errors.
Researchers introduce DALM, a domain-algebraic language model that structures generation into three phases to reduce interference between different knowledge domains. This method could improve accuracy in specialized applications.
Researchers introduce CoLabScience, a proactive AI assistant designed to enhance biomedical collaboration. This innovation addresses the limitations of reactive LLMs by enabling context-aware interventions.
Canada's Federal AI Register, launched in 2025, is more than a transparency tool—it actively shapes accountability. A new study reveals its limitations and biases in tracking AI systems. The register omits key details, raising questions about its effectiveness.
Researchers used Brain Score to evaluate language models trained on diverse languages and structured sequences, finding shared processing properties. The study suggests neural models capture universal linguistic features beyond specific language structures.
Researchers propose a novel approach to optimize LLM agent skills using Monte Carlo Tree Search. This method could significantly improve task performance by systematically refining skill structures and content.
Web Agent Bridge introduces an open-source operating system for AI agents, combining MIT and Open Core licensing. It aims to standardize AI agent development and deployment.
UmaBot is an open-source framework for building multi-agent AI assistants. It leverages LangChain and other tools to enable complex, collaborative AI workflows.
Uber's CTO admits the company is struggling to justify further AI investments despite spending $3.4 billion. The slowdown highlights the challenges of scaling AI initiatives in a tight economic climate.