Live intelligence

Latest AI updates

650 updates

Research

Domain-Specific LLMs Improve Clinical Trial Outcome Prediction, Probing Matches Fine-Tuning

Researchers evaluated BioBERT and BioMistral for phase-level clinical trial outcome prediction on the TOP benchmark, converting trial registrations into text from ten ClinicalTrials.gov fields. All four language-model variants outperformed the structured HINT baseline in ROC-AUC and PR-AUC, with relative PR-AUC gains of 10-20%. Probing with frozen embeddings performed comparably to fine-tuning, suggesting a computationally cheaper alternative for this task.

arxiv4d agoScore 78

clinical trial outcome predictiondomain-specific language modelsefficient fine-tuning

Research Editor selected

Simple Features Outperform Most AI Control Monitors on Standard Dataset

A new arXiv paper shows that simple stylistic and problem-level features can outperform 20 of 24 open-weight AI control monitors on a standard backdoor detection dataset, without executing code. The dataset leaks labels through author style and problem characteristics, and one in seven backdoors fails validity checks, distorting comparisons. The authors release a rebuilt dataset and checks, cautioning that high scores on such benchmarks do not prove sabotage detection.

Why it matters: This research exposes critical flaws in AI safety evaluation benchmarks, offering actionable insights for improving control monitors and dataset design.

arxiv4d agoScore 85

AI safetybenchmark designmodel evaluation

Research Editor selected

Latent-Lens: Enabling Visual Perception in Small Language Models via Latent Communication

Latent-Lens is a lightweight codec that lets a frozen vision-language model (VLM) pass visual information directly to a frozen small language model (SLM) through hidden states, bypassing the lossy text bottleneck. It trains only about 1.6 million parameters (0.2% of combined model size) and improves visual question answering and image captioning over text-based communication with the same LoRA budget. The method is validated across different model architectures and sizes.

Why it matters: This paper introduces a novel, parameter-efficient method for cross-modal latent communication that significantly boosts SLM performance on visual tasks, offering practical value for modular AI systems.

arxiv4d agoScore 82

Latent-LensVLM-SLM bridgingefficient fine-tuning

Research

ERIM: Explicit Rule Induction for Multimodal In-Context Learning

Researchers propose ERIM, a framework that improves multimodal in-context learning by explicitly inducing latent rules from demonstrations rather than relying on direct input-output mapping. It uses a two-stage training strategy combining supervised fine-tuning and reinforcement learning to teach MLLMs to perform rule induction. Experiments show consistent performance gains and improved stability across diverse MICL settings.

arxiv4d agoScore 78

multimodal learningin-context learninglarge language models

Research

Clinical-Protocol-Aware Cyber Resilience Framework for Connected Medical Devices

This paper introduces a cyber-resilience framework for connected clinical technologies that integrates clinical protocols into security design and evaluation. Using Monte Carlo simulations, it shows that segmented architectures reduce cyber exposure by 56.3% compared to connected ones, and combined prevention and containment cut exposure by 60.1%. The study also reveals that technical cyber exposure and clinical impact can diverge, highlighting the need for clinical-context-aware security controls.

arxiv4d agoScore 72

cybersecurityhealthcare technologyAI-enabled threats

Research

CoFailMed: Auditing Correlated Failures in Frozen-Encoder Pools for Medical Segmentation

A new arXiv study introduces CoFailMed, an audit of correlated failures in pools of pretrained encoders for medical image segmentation. Across five 2D tasks, the authors evaluate 11 pretrained encoders and find that shared-checkpoint pairs exhibit the highest failure correlation, and that distinct checkpoints from the same encoder family remain more correlated than cross-family pairs. The study shows that pooling diverse encoders yields effective sizes as low as 1.56-2.92 at a 33rd-percentile Dice threshold, and that fine-tuning does not consistently increase effective pool size. The authors recommend reporting absolute dependence and threshold-specific effective size alongside accuracy and nominal pool size.

arxiv4d agoScore 78

CoFailMedfrozen-encoder poolsmedical image segmentation

Research

Implicit Assumptions About Tasks in Continual Learning

This paper clarifies that continual learning measurements depend on factors like data similarity, change rate, and order, which can be defined at different levels: underlying data, partition, or learner. The authors show these levels often don't coincide and are usually left implicit, leading to ambiguous results. Through teacher-student experiments, they demonstrate that changing granularity, order, or similarity structure alters measured outcomes. They also review benchmarks to see which properties are stated, noting that for foundation models with undisclosed data, properties are inherited by the model's initial state.

arxiv4d agoScore 72

continual learningevaluation methodologytask design

Research Editor selected

Re3Culture: A Human-Grounded Dataset for Cultural Value Alignment

This paper introduces Re3Culture, a large-scale dataset designed to ground cultural value alignment in real human evidence. It addresses three key questions: whose values, which contexts, and what signals, by collecting data from 700+ participants across four cultures, including value surveys, behavioral principles, and responses to situated questions. The analysis reveals that community affiliation, concrete scenarios, and explicit value elicitation are crucial for accurate cultural value estimation.

Why it matters: Provides a novel, human-grounded resource and empirical insights that directly inform the target definition for cultural AI alignment, a critical open problem.

arxiv4d agoScore 85

cultural value alignmentdataset constructionAI alignment

Research

LYNX-ECG: Sparse Visibility Graph Neural Networks for Accurate and Auditable ECG Classification

Researchers introduce LYNX-ECG, a graph neural network that represents heartbeats as sparse visibility graphs of timestamped turning points for ECG classification. On CODE-Test, it achieves a macro F1 score of 0.889, outperforming five baselines including the ECGFounder foundation model, while using only about 14% of the storage of full-sample graphs. The model enables clinical auditing by deleting waveform regions or altering beat timing, with explanations matching clinical expectations in over 90% of correct conduction disorder predictions.

arxiv4d agoScore 78

LYNX-ECGvisibility graphsECG classification

Research

Backpropagation-free Learning for Time-Series Classification Lags Behind Standard Methods

A systematic study evaluates three backpropagation-free learning rules (DFA, DRTP, NoProp) on 112 UCR time-series datasets with 1D convolutional architectures. The optimized learners achieve mean F1 scores of 0.76–0.79, below ROCKET (0.83) and InceptionTime (0.84). DRTP reduces training FLOPs by 33% and speeds up only for long series, while NoProp cuts peak memory by 38% at the cost of longer training.

arxiv4d agoScore 72

time-series classificationdeep learning trainingbackpropagation alternatives

Research

RECAST: Reconstructing Chain-of-Thought from Closed Reasoning Models Without Hidden Channels

RECAST is a new open-source framework that reconstructs chain-of-thought (CoT) reasoning from closed-source models like Claude-Sonnet-5.5, GPT-6.1, and Gemini 3 using only the final answer. It addresses the structural failure of existing extraction methods, which rely on hidden reasoning fields or side channels that are absent in standard chat interfaces. The system conditions a decoder on the question and final answer to produce structured rationales, then cleans them for SFT-ready data. The preprint includes a threat model and controlled artifact validation but does not claim benchmark improvements.

arxiv4d agoScore 78

chain-of-thought extractionreasoning modelsmodel transparency

Research

ReuScope: Auditing Data Reuse and Evaluation Sensitivity in Capsule Vision 2024

ReuScope is a protocol for auditing data reuse and evaluation sensitivity, applied to the Capsule Vision 2024 challenge. It reveals byte-identical files crossing the public training-validation boundary and shared source recordings between partitions. Removing Kvasir-Capsule training rows drops public-validation balanced accuracy from 82.5% to 61.2%, but the effect reverses on other sources, indicating sensitivity rather than leakage inflation. The protocol distinguishes observed reuse from conditional performance sensitivity with matched controls and uncertainty intervals.

arxiv4d agoScore 72

data reuseevaluation methodologymedical imaging

Research Editor selected

Dynamical Memory Transformer: Selective Recurrent Memories with Constant Cost Attention

Researchers present the Dynamical Memory Transformer (DMT), a hybrid architecture that separates sequence encoding from information reading. Several memory units, each a dynamical system with a complex diagonal recurrent matrix and an input-dependent leak rate, are read by attention without going through token history. The recurrence supports an associative scan with O(T) total work and O(log T) parallel depth, while recurrent states and the convolution cache remain fixed-size during autoregressive inference. Experiments show gains on several cognitive tasks and stronger length extrapolation on selective copy, with language-modeling performance close to Transformer but behind Mamba.

Why it matters: It offers a concrete architectural alternative for constant-cost long-context inference, with candid comparisons against Transformer and Mamba.

arxiv4d agoScore 82

Dynamical Memory TransformerEfficient attentionLong-context modeling

Research

SimXAI: A Lightweight LLM-Centric Layer Improves Conversational XAI

SimXAI replaces the rigid GUS-style modules of conversational explainable AI agents with a single tool-calling LLM that bridges to existing XAI resources. It achieves strong function-selection accuracy and high faithfulness while better handling context-dependent follow-up questions. In a user study replacing only the conversational layer of TalkToModel, users completed tasks more accurately and in fewer turns.

arxiv4d agoScore 78

explainable AIconversational agentsLLM integration

Research

LIT: Co-Evolving Video Director Agents Improve Production and Interaction

The paper introduces LIT, a framework for video director agents that co-evolves directing and user interaction capabilities. It uses a simulated user to probe weaknesses and a meta engineer to revise strategies, forming an evolutionary loop. On Test and OOD Test splits, LIT improves production and interaction scores by 19.1% and 32.0% relative to the initial pipeline.

arxiv4d agoScore 72

AI agent evolutionvideo generationhuman-AI interaction

Research Editor selected

AI evaluators favor job applications written by certain AI models, independent of candidate quality

A new study introduces a multi-model auditing framework that pairs eleven LLM writer agents with nine evaluator agents across ten job descriptions and fifty candidates per job. It finds a consistent generative preference hierarchy where locally hosted models are systematically disadvantaged, compounded by self-preference bias. In the most adverse pairing, about 1.8 out of 25 qualified candidates are filtered out purely due to model mismatch. The study offers actionable guidance for fairer screening design and discusses implications for algorithmic fairness and regulation.

Why it matters: This research provides concrete evidence of AI model bias in hiring, a high-impact area with practical implications for fairness and regulation.

arxiv4d agoScore 85

algorithmic fairnessAI-assisted screeningLLM evaluation

Research

Automated Keyword Generation Using LLMs in a High-Volume Data Virtualization Service

This paper presents a solution for automatically generating keywords for virtual datasets in a data virtualization service (DVS) using large language models (LLMs). The authors propose five strategies for selecting text data for LLM processing and evaluate them across three LLM generators and two embedding techniques in five application case studies. Results show that strategy performance depends on text description quality, leading to a multi-strategy implementation in a new keyword generation service. The work highlights the importance of automated keyword generation for dataset-rich, frequently accessed DVS environments.

arxiv4d agoScore 72

large language modelsdata managementautomated metadata generation

Research Editor selected

General-purpose foundation models outperform specialized chemical FMs on property prediction with scarce data

A new arXiv study compares specialized chemical foundation models (ChemBERTa, MoLFormer) with general-purpose models (GPT-6 Astra, Claude Opus 5.5) on chemical property prediction tasks. With only 60 experimental datapoints, GPT-6 Astra outperforms the chemical FMs on five regression tasks. Chemical FMs only match Astra's performance when given 10x more data, suggesting general-purpose models are better for data-scarce chemistry research.

Why it matters: This finding challenges the common assumption that domain-specific models are superior in specialized fields, offering practical guidance for researchers in chemistry and AI.

arxiv4d agoScore 82

foundation modelschemistryproperty prediction