Skip to main content
Search

Hong Kong Baptist University

5 related articles

Subscribe to updates
  1. arXiv

    SafeActBench traces the evidence-to-action chain across 656 cases, finding strong static judgment but weaker interactive execution in tool-using agents

    The authors introduce SafeActBench, 656 cases across six operational domains and five protocols, using a provenance-bound Evidence Ledger and a deterministic trajectory evaluator to check whether tool-using agents establish required evidence before acting; across ten model–harness configurations, strong static action assessment coexists with much weaker interactive execution, failures often begin with incomplete investigation or premature action, single-action execution is usually reliable once evidence is established, and multi-action workflows additionally expose unresolved prerequisites and incomplete execution.

  2. arXiv

    HarnessSecurity-Bench tests six coding agents: enabling auto-approve raises attack success from 29.2% to 95.6%

    The work builds a ten-mechanism taxonomy across 40 coding agent harnesses and rates 400 harness–mechanism cells (205 confirmed implemented, 83 confirmed absent, 112 unresolved), then introduces HarnessSecurity-Bench, which runs 2,500 trials under GLM-5.2 on 23 tasks across five attack surfaces with separate deterministic oracles for six harnesses, finding that enabling auto-approve raises attack success from 29.2% to 95.6%, that network isolation and read-only mode cut attack success from 57.1% and 8.3% to 0.8% at utility losses of 24.5 and 34.0 percentage points, and that command allowlisting and denylisting reduce attack effects with a small utility loss and a utility gain respectively.

  3. arXiv

    TTCL jointly improves LLM reasoning accuracy and confidence calibration on unlabeled target data, with +40.13% average relative accuracy and +70.80% ECE reduction on base models

    The work proposes Test-Time Calibration Learning (TTCL), a label-free framework that derives self-supervision signals for correctness and calibration from multiple model-generated responses and jointly adapts reasoning accuracy and verbalized confidence on unlabeled target-task data, reporting consistent improvements in both accuracy and calibration on mathematical reasoning and factual question answering.

  4. arXiv

    KGPFN uses in-context learning to give knowledge graph foundation models the best average MRR across 57 graphs

    The authors propose KGPFN, a knowledge graph foundation model built on a Prior-Data Fitted Network that learns relation representations via message passing on relation graphs, extracts multi-scale local context from the intermediate head representations of a multi-layer NBFNet, and builds relation-specific global context from positive and negative examples of the query relation, aggregating this context with feature-level and sample-level attention; after multi-graph pretraining it combines structural representations with labeled contextual evidence without inference-time parameter updates, achieving the best average MRR on 57 knowledge graphs both without and with fine-tuning, with context sensitivity analyses highlighting the value of negative context examples.

  5. arXiv

    GenFT: A Generative Parameter-Efficient Fine-Tuning Method for Pretrained Foundation Models

    This work proposes GenFT, a W0-conditioned parameter-efficient fine-tuning method in which a deterministic generator produces task-specific updates ΔW by applying row and column transformations to the pretrained weights W0 together with a shared-specific decomposition, reporting competitive or better average performance on GLUE (RoBERTaBase, 85.87% average with 0.24M parameters), VTAB-1K (ViT-B/16, 74.50% average with 0.27M parameters) and FGVC (90.38% average with 0.29M parameters), plus a perplexity pilot study on LLaMA-7B with an Alpaca subset.

You’re all caught up