User modeling in industrial recommender systems typically produces dense embeddings, which suffer from representational constraints due to fixed-dimensional vectors. As an emerging alternative for discrete user representation, LLM-generated natural-language user profiles often capture language semantics rather than seq...
In open large language model (LLM) ecosystems, models are frequently adapted across multiple domains and applications, forming multi-stage derivation chains. Consequently, tracking and verifying historical contributions is essential for model provenance and intellectual property protection. However, existing watermarki...
While transcoder-based circuit tracing has recently enabled detailed causal analyses of large language models, it remains unclear how semantic information propagates across denoising steps and how text and image representations interact within multimodal diffusion transformers (MM-DiT) for image generation. Attention m...
ST-AudioQA, a spatio-temporal audio QA dataset and benchmark built from first-order ambisonic (FOA) renderings of static and moving sound sources, and ST-Audio Encoder, which jointly learns event semantics and time-resolved source trajectories, and ST-AudioLM, which exposes these audio tokens to an LLM for spatio-tempo...
Video world models are increasingly used to provide predictive visual representations, yet it remains unclear which pretraining signals induce action-relevant structure in their latent spaces. We study this question through a unified probe-based evaluation across diverse encoder families, including image-only self-supe...
Jewon Yeom, Hanseul Kim, Jeongjae Park et al.· 0 citations
Large language models can reproduce training data, but memorization evaluations mostly measure whether models can be made to do so, not whether they do so without adversarial prompting. We introduce PropMe, a framework that pairs a capability evaluation (prefix attacks) with non-adversarial propensity evaluation, and a...
Gianluca Barmina, Peter Schneider-Kamp, Lukas Galke Poech· 0 citations
Many modern vision-language models (VLMs) build on autoregressive decoding of discrete tokens. While text-based output interfaces enable scalable pretraining and strong zero-shot generalization across diverse tasks, they are poorly suited for problems that require precise continuous outputs, such as localizing temporal...
Zhuoming Liu, Jinhong Lin, Kwan Man Cheng et al.· 0 citations
Multi-agent reasoning systems adopt a generate-then-transfer paradigm that forces end-to-end latency to scale linearly with pipeline depth. We introduce StreamMA, a multi-agent reasoning system that streams each reasoning step to downstream agents as soon as it is generated, pipelining adjacent agents and thus reducing...
Zhen Yang, Xiaogang Xu, Wen Wang et al.· 0 citations
In robotic manipulation, the tight coupling between grasping and motion planning often obscures the true source of failure, leading to inefficient trial-and-error. To enable efficient long-horizon manipulation, we propose GTP-FA (Grasp-Then-Plan with Failure Attribution), a task-oriented two-stage 'grasp-then-plan' fra...
Jiahao Xu, Peiyuan Wang, Hanzhuo Zhang et al.· 0 citations
Reliable probability estimates are critical in many machine learning applications, yet modern classifiers are often poorly calibrated. Post-hoc calibration provides a simple and widely used solution, but the large number of proposed methods, combined with small-scale and inconsistent evaluations, makes it difficult to...
Eug\`ene Berta, David Holzm\"uller, Francis Bach et al.· 0 citations
PlanAhead is introduced, a static planner-executor framework that evaluates the impact of plan representation on agent performance and shows that both, the plan formulation and the underlying LLM generating the plan, significantly influence web-agent robustness and task success.
Alejandra Zambrano, Sara Vera Marjanovic, Imene Kerboua et al.· arXiv.org· 1 citation
Of $1{,}536$ Gaussian release covariances we tested for single-layer hidden-state privacy, zero achieve both moderate utility and moderate privacy against an adaptive retrieval attacker. We prove a complementary Fisher-ball lower bound: every full-rank Gaussian release at $O(1)$ Fisher utility admits a direction whose...
