Skip to main content
Search

Vector Institute

5 related articles

Subscribe to updates
  1. arXiv

    SPARC lets experimental evidence rewrite model weights: 8 of 11 recommended conditions fall below 300 mV overpotential

    Closed-loop experiments usually fix the model family and representation first, then let new data update only predictions. SPARC builds an evidence matrix from the out-of-sample predictions of competing regression models, assigns each model's influence through a sparse posterior, and combines candidate-dependent predictive dispersion with a residual scale into a trust score that modulates exploration. In an automated alkaline oxygen-evolution campaign, 8 of 11 recommendations fixed in advance reached overpotentials below 300 mV, versus 18 of 125 initial experiments, with a best value of 242 mV.

  2. arXiv

    Prefill jailbreaks meet a training-free fix: restore the refusal signal the attack erodes, request by request

    Prefill jailbreaks force an unsafe opening into the assistant turn so a model that would refuse continues instead. Tracking vocabulary-grounded signals with the Jacobian lens shows refusal weakening faster than harmfulness as the forced prefix deepens, reaching the benign baseline by free generation. J-Anchor, a training-free inference-time defense, anchors each request to its pre-attack assistant-header state and restores only the refusal components that fall below that target during prefilling and decoding, cutting attack success on fixed and deep prefills across four open-weight instruction models while keeping benign refusal and reasoning close to the undefended model.

  3. arXiv

    MemCo pairs local and global memory so LLM agents gain 43.47% held-out ALFWorld accuracy and 32.10% fewer steps in unseen environments

    The authors propose MemCo, a memory-centric collaboration framework in which each agent keeps an environment-specific local memory graph while evidence-aggregated, Wilson-lower-bound-filtered transferable workflows are promoted to a shared global memory, and at decision time local and global memories are retrieved and grounded by current state, task phase, and goal; across ALFWorld, PDDL, FEVER, and ScienceWorld with multiple LLM backbones, MemCo attains the highest accuracy in 10 of 12 evaluation settings and the fewest interaction steps in 9, e.g., on Qwen3-4B it improves ALFWorld held-out layout accuracy by 43.47% and reduces steps by 32.10% relative to the runner-up baseline.

  4. arXiv

    DUCX decomposes chest X-ray agent unfairness into tool exposure, tool transition, and reasoning, finding gaps up to about 50% beyond end-to-end metrics

    Using MedRAX as the instantiated system, this work audits fairness in tool-using chest X-ray question-answering agents and proposes DUCX, a stage-wise decomposition that separates end-to-end bias into tool-exposure bias, tool-transition bias, and LLM reasoning bias; across five driver LLMs on CheXAgentBench and the curated MIMIC-FairnessVQA, demographic gaps persist end to end (equalized odds up to 20.79%, lowest fairness-utility tradeoff down to 28.65%), and subgroup disparities in tool usage, routing patterns, and reasoning traces are not predictable from end-to-end evaluation alone (for example, conditioned on segmentation-tool availability the subgroup utility gap reaches as high as 50%).

  5. arXiv

    Open-PMC-18M builds 18M medical image-text pairs via subfigure splitting and context summaries, lifting average retrieval by 27%

    Starting from the BIOMEDICA corpus, the authors used DAB-DETR-based subfigure detection, Qwen2.5-VL-32B-Instruct subcaption extraction, and Qwen2.5-14B-Instruct summarization of inline context to curate and release OPEN-PMC-18M, a dataset of 18 million subfigure-text pairs, then trained vision-language encoders evaluated on 6 retrieval and 19 zero-shot classification tasks across radiology, microscopy, and visible light photography, reporting an average recall of 21.64, a 27% relative improvement over the previous best model OPEN-PMC, and an overall zero-shot average F1 of 39.77.

You’re all caught up