1035 stories curated by AInformed · page 43 of 44
New research reveals that safety-trained language models routinely refuse requests to help users evade unjust, absurd, or illegitimate rules. This phenomenon, termed 'blind refusal,' highlights a critical gap in AI moral reasoning where compliance overrides ethical judgment.
BDI-Kit introduces a dual-interface toolkit for data harmonization, combining Python APIs for developers and AI chat interfaces for domain experts. This approach addresses the longstanding challenge of integrating disparate datasets.
Researchers developed machine learning models to predict container service needs and dwell times, reducing unproductive moves at terminals. The study leverages historical data to improve operational efficiency in shipping logistics.
Researchers analyzed 1,108 audio-recorded primary care visits to train AI models that detect depression from naturalistic dialogue. The best-performing model, combining Sentence-BERT with logistic regression, achieved high accuracy in identifying patients with PHQ-9 confirmed depression.
Researchers introduce AgentGate, a new routing engine designed to solve dispatch inefficiencies in the emerging Internet of Agents. By replacing unrestricted text generation with structured candidate-aware routing, it optimizes latency, privacy, and cost.
Researchers introduce ADAG, a new pipeline that automatically describes attribution graphs in language models, eliminating the need for manual circuit tracing. This shift promises to scale interpretability research by replacing ad-hoc human inspection with automated analysis.
Researchers propose a weak supervision framework to detect hallucinations in large language models. This approach enables hallucination detection from internal activations alone at inference time.
Researchers propose a new approach to sequential clinical diagnosis using uncertainty-guided latent diagnostic trajectory learning. This method addresses the challenge of learning effective diagnostic trajectories under uncertainty.
Researchers estimated the state-space complexity of Shogi using the Monte Carlo method. The study aims to determine the number of reachable positions in the game.
SELFDOUBT is a new framework for uncertainty quantification in reasoning language models. It addresses the difficulty of deploying uncertainty estimation in practice, particularly for proprietary APIs.
ReVEL is a hybrid framework that uses large language models for iterative reasoning in combinatorial optimization. It embeds LLMs within an evolutionary algorithm to improve heuristic design.
Researchers challenge the notion that supervised finetuning memorizes while reinforcement learning generalizes. They find that cross-domain generalization is conditional, influenced by optimization, data, and model capability. This challenges prevailing narratives in LLM post-training.
Large reasoning models perform well on multi-step tasks but have unstable behavior. Step-Saliency analysis reveals information-flow failures.
ProofSketcher combines LLMs with a lightweight proof checker for reliable math and logic reasoning. It aims to address the limitations of LLMs in producing persuasive but flawed arguments.
Pramana is a novel approach to fine-tune large language models for epistemic reasoning. It aims to address the epistemic gap in AI, where models struggle with systematic reasoning and often produce unfounded claims.
Pramana is a novel approach that teaches large language models explicit epistemological methods to improve their reasoning. This approach aims to address the epistemic gap in AI, where models struggle with systematic reasoning and often produce unfounded claims.
Pramana is a novel approach to fine-tune large language models for epistemic reasoning. It aims to address the epistemic gap in LLMs, enabling them to ground claims in traceable evidence.
PaperOrchestra is a multi-agent framework that automates AI research paper writing. It transforms unstructured materials into submission-ready manuscripts, including literature synthesis and generated visuals.
Researchers introduce a framework to study operational noncommutativity in sequential metacognitive judgments. This work explores how order effects impact cognitive processes.
MMORF is a framework for designing multi-objective retrosynthesis planning systems. It leverages language model-based multi-agent systems to balance quality, safety, and cost objectives.
A new mathematical theory models the evolution of self-designing AIs. This theory differs from biological evolution due to directed descendant design.
Researchers propose a mathematical theory to model the evolution of self-designing AIs. This theory aims to understand how AI systems shape their descendants through recursive self-improvement.
ATANT is an open evaluation framework for measuring continuity in AI systems. It assesses the ability to persist and update meaningful context across time.
Researchers propose a framework to uncover algebraic structures in combinatorial optimisation problems. This approach reduces search spaces and improves global optimal solution discovery.