Skip to main content
Search

Technische Universität Darmstadt

7 related articles

Subscribe to updates
  1. Fact-checking articles can now be drafted automatically: QRAFT chains evidence, drafting and simulated editing, but experts still want a human in the loop

    QRAFT splits fact-checking article writing into evidence gathering, drafting and a simulated editorial review run by multiple LLM agents, beating prior text-generation methods on automatic metrics across 1,227 examples while still scoring well below expert-written articles with professional fact-checkers.

  2. arXiv

    SpectralCache reuses stable singular subspaces and extrapolates singular values to reach 5.22x acceleration on HunyuanWorld-Voyager-13B while holding a WorldScore of 65.90

    The work reveals that diffusion-based world-model features have highly stable singular subspaces across nearby denoising steps with predictable singular-value evolution, and builds on this to propose SpectralCache, a training-free spectral caching framework that reuses stable singular subspaces and estimates only low-dimensional singular values via linear extrapolation, while exploiting spectral consistency between neighboring full-computation features to skip selected expensive backbone evaluations via singular value scaling; on HunyuanWorld-Voyager-13B it achieves 5.22x acceleration while maintaining a WorldScore of 65.90 for static scenes, substantially outperforming existing training-free caching methods in inference efficiency.

  3. arXiv

    AutoDataBench scores Kimi-3 highest at 60.68 overall, yet the retrieval leader on in-distribution data ranks last out of distribution

    The work introduces AutoDataBench, a controlled testbed that isolates data intelligence under fixed training pipelines and resource budgets across tool-use data diagnosis and repair, retrieval data organization, and knowledge injection, evaluates seven frontier LLMs on iterative data decisions, finds that agent-curated data approaches expert references on tool use and retrieval while strong in-distribution scores do not guarantee out-of-distribution transfer, and shows that reusing optimization trajectories as mid-training data improves downstream coding performance.

  4. arXiv

    LLMLeak turns an LLM's web-fetch tool into a covert channel, reaching a 79.7% attack success rate across eleven open-parameter models

    The paper introduces LLMLeak, a novel attack vector in which malicious software that runs locally but cannot reach the internet directly encodes a secret into a URL and presents the referenced site as information needed for a benign task such as migrating a software library, inducing the LLM to fetch that URL through its web-fetch tool so the attacker receives the encoded secret via an attacker-controlled DNS or web server; evaluation on eleven open-parameter models shows a 79.7% attack success rate, with a case study on real-world chatbots.

  5. arXiv

    SnapLog turns kitchen videos into event logs via ViT frame embeddings, K-Means segmentation and R(2+1)D few-shot classification, reaching 90% and 85% Top-3

    The work presents SnapLog, a pipeline that encodes each video frame with a pre-trained ViT, performs temporal segmentation through a frame-wise similarity matrix with K-Means and a greedy merge, then applies generalized few-shot classification with a 34-layer R(2+1)D encoder pre-trained on Sports-1M plus a linear head to convert raw video into timestamped event logs (either deterministic or uncertainty-aware logs that retain label probability distributions); on 16 brownie-baking videos from TUM Kitchen and 24 kitchen-cleaning videos from Epic Kitchens-100 it achieves frame-wise segmentation accuracy of 74.3% and 67.7%, and with augmentation plus 100 overlapping clips Top-3 classification accuracy of 90.1% and 85.

  6. arXiv

    TUM team surveys 329 works on physics-embedded robot learning: 71% encode physics in architectures, only 4% combine multiple embedding routes

    This survey systematically reviews methods that embed physics priors into robot learning and proposes a unified taxonomy classifying existing work by where physics is embedded—physics-guided inputs, data, and representations; physics-encoded model architectures; and physics-informed training losses—then reviews methods for robot dynamics learning, trajectory planning and prediction, control, and estimation together with the open-source software ecosystem, reporting that physics-encoded architectures account for 71% of surveyed methods, physics-guided 16%, and physics-informed 13%, while only 4% combine more than one embedding route.

  7. Science Advances

    Neural ODEs and symbolic regression recover an ice-crystal growth equation from 290 mass time series and reproduce early growth in independent AIDA chamber data

    Using neural ordinary differential equations (NODEs) optimized simultaneously across 290 mass-ratio time series of ice crystals grown in a levitation diffusion chamber, the study learned the unknown functional form of the transfer coefficient G in the depositional ice growth model, then used symbolic regression (PySR) to derive a closed-form expression; the weakly-constrained NODE model performed best on 138 of 290 experiments (MSE loss 16495, versus 41396 for no surface kinetics and 40751 for Nelson and Baker 1996), yielding G = a0·Gca1·[a2 + a3/m]^-1 + a4 with an added ice-crystal-mass term, and this expression predicted ice supersaturation and ice water content better than the capacitance growth model and Nelson and Baker 1996 in independent AIDA Aerosol and Cloud Chamber experiments.

You’re all caught up