261 stories tagged LLMs · page 7 of 11
Researchers propose a structured approach to debugging LLMs, treating them as observable systems. The method offers model-agnostic techniques for issue detection and refinement, addressing the complexity of LLM errors.
Researchers propose a new AI-driven framework for aircraft fault diagnosis using digital twins and LLMs. The method addresses data scarcity and diverse fault types in general aviation. The system integrates high-fidelity simulations, fault injection, and interpretable report generation to improve diagnostic accuracy.
IBM has released Granite 4.1, a suite of open-source large language models trained using advanced techniques like mixture-of-experts and fine-tuning. The models are optimized for both efficiency and performance, setting a new standard in open-source AI.
Researchers introduce FormalScience, an AI pipeline that helps domain experts formalize scientific reasoning into verifiable code. This addresses a key challenge in applying LLMs to scientific domains with complex notations.
A new study shows that belief graphs significantly improve weaker LLMs in cooperative reasoning tasks, while stronger models see minimal benefits. The research highlights the importance of integration architecture in leveraging these graphs effectively.
Researchers found that the variability in LLM responses to different prompt styles stems from shared underlying task representations. This explains why the same question can yield different answers depending on phrasing.
A new study formalizes the concept of 'background temperature' to describe implementation-level nondeterminism in large language models. This helps explain why identical inputs can produce divergent outputs even at temperature T=0.
A study reveals that LLMs trained on Western data fail to detect health misinformation in non-Western contexts, such as cow urine remedies on Indian YouTube. This highlights a critical gap in AI's ability to handle culturally nuanced content.
Researchers introduce a lightweight retrieval-augmented generation (RAG) framework to improve patient-trial matching. The approach balances scalability and efficiency with the ability to handle complex EHR data and eligibility criteria.
Researchers developed a control-theoretic framework to determine when iterative self-correction improves LLM performance. The study introduces a Markov model diagnostic to assess whether repeated refinement helps or hurts accuracy.
Researchers introduce SHAPE, a new benchmark to evaluate educational LLMs under adversarial conditions. The study highlights 'pedagogical jailbreaks' where students manipulate LLMs to provide answers instead of learning guidance.
A new taxonomy identifies risks like deception and reward hacking in advanced AI systems. The framework aims to benchmark these behaviors as models grow more capable.
A new AI usage analytics tool acts as a proxy to enforce budget limits and redact personally identifiable information before requests reach the provider. It offers real-time cost tracking and immediate suspension when thresholds are breached.
A new paper highlights the risks of AI agents producing selectively chosen, publishable analyses that lack rigorous validation. The study calls for adversarial experiments to ensure scientific integrity.
Researchers developed an AI system that can reproduce social science findings using only paper descriptions and raw data. The method achieves deterministic, cell-level result comparisons without access to original code or outputs.
Routiium is a new self-hosted, OpenAI-compatible LLM gateway that includes a unique tool-result guard feature. This innovation addresses a critical security gap in LLM agent loops by monitoring tool outputs, not just user inputs.
LLM-Rosetta is an open-source tool that standardizes API calls across OpenAI, Anthropic, and Google's large language models. It simplifies integration for developers by abstracting provider-specific differences.
Researchers introduce COMPASS, a system that automates prompt engineering for generating human-understandable explanations of AI task planning. The tool addresses the critical need for reliable, stakeholder-specific explanations in complex software systems.
Researchers developed a novel approach for LLMs to co-evolve decision-making and skill banks, significantly improving performance in complex, long-horizon game environments. This method addresses key challenges in multi-step reasoning and delayed rewards.
ThermoQA is a new benchmark for evaluating LLMs on engineering thermodynamics problems. It shows significant performance drops as problem complexity increases, with top models like Claude Opus 4.6 leading at 94.1%.
A comprehensive guide for running large language models on a 64GB RAM device has been released. It covers practical tips for optimizing performance in code and math applications.
Graeme (@gkisokay) shares a curated list of powerful local LLMs that run efficiently on 32GB RAM machines. This opens up flagship-class models to a wider range of users.
Hugging Face introduces QIMMA, a quality-focused leaderboard for Arabic LLMs. It aims to highlight models that excel in both performance and cultural relevance.
Researchers have identified a pervasive phenomenon called 'tool overuse' in large language models, where they unnecessarily rely on external tools instead of internal knowledge. The study explores the underlying mechanisms behind this behavior, highlighting a 'knowledge epistemic illusion' where models misjudge their own capabilities.