Skip to main content
Search

German Research Centre for Artificial Intelligence

9 related articles

Subscribe to updates
  1. Diffusion language models learn to count characters: no fine-tuning, summaries land within 3.1 characters of target

    For summarization settings such as headlines and previews that must fit an exact character count, Delta estimates the expected character length of the current output inside a masked diffusion language model's denoising loop and dynamically inserts or removes masked positions to steer generation toward the target. It needs no fine-tuning and applies to off-the-shelf LLaDA and Dream. Across the XSum, CNN/DailyMail and arXiv summarization benchmarks, character-length error falls by an order of magnitude relative to the baselines while summary quality stays competitive under automatic and human evaluation.

  2. bioRxiv

    Single-cell perturbation prediction drops the pretraining crutch: GFM wires knowledge graphs into flow matching and trains up to 30x faster

    Across several large Perturb-seq datasets, a generative framework called Graph-guided Flow Matching (GFM) combines three biological knowledge graphs, covering Gene Ontology, protein interactions and perturbation similarity, with latent flow matching to predict single-cell transcriptomes after unseen perturbations. It outperforms a linear baseline and the deep-learning models CellFlow, GEARS and scGPT on unseen single-gene perturbations, trains up to 30 times faster, and keeps its advantage as training data shrink; on two-gene perturbations it does better when neither gene was seen in training.

  3. medRxiv

    Therapy sessions can now run in silico: raising therapist empathy prompts lifts every motivational-interviewing integrity score

    Psychotherapy research is constrained by costly trials and clinical features that are hard to manipulate at random. A preregistered proof-of-concept study paired therapist and patient agents built on a locally deployed Llama-3.3-70B-Instruct to generate 512 automated motivational interviewing sessions, manipulating only the therapist prompts for empathy, affirmation and alliance building. Among 480 sessions retained for analysis, all four motivational interviewing treatment integrity global ratings were higher under high common factors, with adjusted mean differences of 0.54 to 0.95 points.

  4. arXiv

    Visual-head discovery scores shift with where attention is read: necessity tracks the query span, while steering at default strength is matched by random heads

    The internal choices inside a visual-head discovery score reshape which heads it returns in vision-language models. On Qwen3-VL-8B-Instruct, fully crossing query type, discovery corpus and the token positions whose attention is read, and testing all 12 head-sets with steerability and necessity interventions, only necessity follows the attention source: query-span head-sets are more necessary than last-token ones, and the two referring plus query-span sets lie 17–19 standard deviations above 250 layer-matched random sets. At the default strength, no discovered head-set steers better than its layer-matched random sets.

  5. CleanSurvival uses reinforcement learning to auto-select preprocessing pipelines for survival analysis, improving predictive performance over simple baselines on real-world benchmarks

    The work presents CleanSurvival, a reinforcement-learning (Q-learning) based automated data-preprocessing framework tailored to time-to-event (survival analysis) models for censored data: it selects among combinations of data imputation, outlier detection, and feature extraction techniques to optimize performance for a Cox, random forest, neural network, or user-supplied time-to-event model; on real-world dataset benchmarks, this Q-learning-based preprocessing improved predictive performance relative to simple baselines, runtime behavior was condition-dependent and most clearly interpretable in the best-covered benchmark cells, and a simulation study showed effectiveness across different types and levels of missingness and noise.

  6. arXiv

    EAGER boosts generative event extraction with verifiable-reward reinforcement learning, beating prompting, supervised fine-tuning, and prior RL baselines across seven benchmarks

    The work presents EAGER, a reinforcement learning framework for generative event extraction that combines fine-grained verifiable rewards with Schema-Contrastive Advantage Estimation to alleviate advantage collapse under sparse binary rewards; its reward design explicitly targets structural validity, extraction accuracy, groundedness, coverage, over-generation, and span precision, and across seven benchmark datasets it consistently outperforms prompting, supervised fine-tuning, and prior reinforcement learning baselines, achieving a substantial improvement over the strongest prior method.

  7. Medical Concept Normalization of German Clinical Expressions to SNOMED CT: Domain Embedding Retrieval with LLM Reranking Outperforms LLM-Only

    This study investigates medical concept normalization of short German clinical expressions to SNOMED CT by comparing a direct GPT-5.4 LLM-only approach with a hybrid approach combining medBERT.de bi-encoder embedding retrieval and RAG-based LLM reranking, finding that the LLM-only baseline achieves Recall@1 of 0.235, Recall@3 of 0.297, and Recall@5 of 0.303, while embedding-based retrieval reaches Recall@1 of 0.681, Recall@3 of 0.783, and Recall@5 of 0.812, and adding RAG reranking further improves Recall@1 to 0.771 with Recall@3 and Recall@5 at 0.809 and 0.812.

  8. In-Context Molecular Property Prediction with LLMs: A Blinding Study on Memorization and Knowledge Conflicts

    The study evaluates nine LLM variants across three families (GPT-4.1, GPT-5, Gemini 2.5) on three MoleculeNet datasets (Delaney solubility, Lipophilicity, QM7 atomization energy) using a systematic blinding procedure that iteratively reduces available information, complemented by 0-, 60-, and 1000-shot in-context sample sizes, in order to determine whether molecular property prediction reflects genuine in-context regression or verbatim retrieval of memorized target values, and it adds positive and negative controls for the memorization experiments, structural reference baselines for the multi-shot experiments, and bootstrap confidence intervals for all results; it finds no evidence of verbatim retrieval on these legacy benchmarks and shows that blinding exposes conflicts between pre-traine

  9. arXiv

    Just Leaf It: Accelerating Diffusion Classifiers with Hierarchical Class Pruning

    The work proposes a Hierarchical Diffusion Classifier (HDC) that exploits parent-child label hierarchies: in a pruning stage it traverses the label tree level by level, estimates epsilon-prediction errors with fewer Monte Carlo samples, and keeps only the lowest-error nodes, then runs classical diffusion classification on the surviving leaf nodes, achieving roughly 60% inference speed-up on ImageNet-1K with Stable Diffusion 2.0 (1600s to 650s) or raising per-class accuracy from 64.90% to 65.16% at nearly unchanged runtime.

You’re all caught up