Skip to main content
Search

University of Duisburg-Essen

4 related articles

Subscribe to updates
  1. arXiv

    Safety alignment no longer rests on a few thousand neurons: ABS-Align prunes them during finetuning and cuts some attack success by about 90%

    Safety alignment in open-weight models is highly sparse: pruning a small set of safety-critical neurons or ablating a single refusal direction can make a model produce harmful content. ABS-Align gradually prunes those identified safety-critical neurons during preference finetuning, forcing refusal behavior to be re-established through other parts of the network, and restores the pruned neurons afterward. Across Gemma-2-2B-IT, Llama-3-8B-Instruct, Llama-3.2-3B-Instruct and Qwen-3-4B, attack success generally falls against TwinBreak, NeuroStrike, refusal-direction ablation and ten SORRY-Bench black-box attacks, with reductions of about 90% for some attacks, while TruthfulQA improves and general capabilities change little.

  2. A single-cell velocity model you can perturb: it named three rescue targets for Arid1a-loss B cells, and mice confirmed all three

    DynaVelo puts RNA expression and transcription-factor motif accessibility from single-cell multiome data into one neural differential equation, learning a velocity field that can be perturbed rather than only drawn. In wild-type mouse germinal center B cells it recovers dark-zone cycling, the switch to the light zone and recycling; in Arid1a-heterozygous cells it nominated Ctcf, Stat3 and Bcl6 as rescue targets, and all three double-heterozygous mouse strains restored the dark-zone phenotype.

  3. arXiv

    Among 20 university students, writing with a glove showed no clear effect on immediate recall, while increased writing pressure reduced recall with about 85–88% probability

    Limbu and Chounta used a 2x2 between-subjects design in which 20 right-handed German-native university students copied text on a WACOM tablet while touch sensitivity (glove use) and kinaesthetic intensity (increased writing pressure) were manipulated, measuring outcomes with a 10-item immediate recall test, dual-task reaction time, and NASA-TLX; Bayesian binomial regression showed about 85–88% probability that increased pressure reduced recall (Bayes factors 5.83 and 7.1, moderate evidence), glove use alone showed no clear effect, and Bayesian mediation analysis found no strong evidence that mental effort or perceived workload mediated these effects (all 95% credible intervals included zero).

  4. arXiv

    pm4aa mined eight roles from eight years of Commitizen repository logs and generated five executable AI agents, with smoke tests routing correctly but autonomy scoring lowest in the user study

    The work presents pm4aa, a pipeline that extracts object-centric event logs from GitHub repositories via PyStack't, maps commits to SE tasks with a Conventional Commits regular expression, partitions 589 users into eight roles with a priority-ordered rule classifier, applies object-centric, imperative (BPMN), and declarative (DECLARE) process mining per role, has an LLM generate process descriptions, and synthesizes a LangGraph multi-agent application through IBM BOB; on the Commitizen project (November 2017 to November 2025, 21,488 events, 4,813 objects) it produced five role agents, three smoke tests routed correctly, and a ten-participant user study rated knowledge schema, operational clarity, and human engagement at a median of 4, accountability at 3, and autonomy at only 2.

You’re all caught up