arXiv subjects

Jacob Hansen

Publications and source records attributed to Jacob Hansen.

7 recordsLinked to original sources

arXiv2025

Instructify: Demystifying Metadata to Visual Instruction Tuning Data Conversion

Visual Instruction Tuning (VisIT) data, commonly available as human-assistant conversations with images interleaved in the human turns, are currently the most widespread vehicle for aligning strong LLMs to understand visual inputs, converting them to strong LMMs. While many VisIT datasets are available, most are constructed using ad-hoc techniques developed independently by different groups. They are often poorly documented, lack reproducible code, and rely on paid, closed-source model APIs such as GPT-4, Gemini, or Claude to convert image metadata (labels) into VisIT instructions. This leads to high costs and makes it challenging to scale, enhance quality, or generate VisIT data for new datasets. In this work, we address these challenges and propose an open and unified recipe and approach,~\textbf{\method}, for converting available metadata to VisIT instructions using open LLMs. Our multi-stage \method features an efficient framework for metadata grouping, quality control, data and prompt organization, and conversation sampling. We show that our approach can reproduce or enhance the data quality of available VisIT datasets when applied to the same image data and metadata sources, improving GPT-4 generated VisIT instructions by ~3\% on average and up to 12\% on individual benchmarks using open models, such as Gemma 2 27B and LLaMa 3.1 70B. Additionally, our approach enables effective performance scaling - both in quantity and quality - by enhancing the resulting LMM performance across a wide range of benchmarks. We also analyze the impact of various factors, including conversation format, base model selection, and resampling strategies. Our code, which supports the reproduction of equal or higher-quality VisIT datasets and facilities future metadata-to-VisIT data conversion for niche domains, is released at https://github.com/jacob-hansen/Instructify.

cs.CV↗

arXiv2025

Granite Vision: a lightweight, open-source multimodal model for enterprise Intelligence

We introduce Granite Vision, a lightweight large language model with vision capabilities, specifically designed to excel in enterprise use cases, particularly in visual document understanding. Our model is trained on a comprehensive instruction-following dataset, including document-related tasks, such as content extraction from tables, charts, diagrams, sketches, and infographics, as well as general image tasks. The architecture of Granite Vision is centered around visual modality alignment with a decoder-only, 2 billion parameter Granite large language model. Additionally, we introduce a dedicated safety classification approach in test-time that leverages a sparse set of attention vectors to identify potential harmful inputs. Despite its lightweight architecture, Granite Vision achieves strong results in standard benchmarks related to visual document understanding, as well as on the LiveXiv benchmark, which is designed to avoid test set contamination by using a constantly updated corpus of recently published Arxiv papers. We are releasing the model under the Apache-2 license, allowing for both research and commercial use, while offering complete visibility into the training data and other relevant details. See https://huggingface.co/ibm-granite/ for model weights.

cs.CV↗

arXiv2024

Self-MoE: Towards Compositional Large Language Models with Self-Specialized Experts

We present Self-MoE, an approach that transforms a monolithic LLM into a compositional, modular system of self-specialized experts, named MiXSE (MiXture of Self-specialized Experts). Our approach leverages self-specialization, which constructs expert modules using self-generated synthetic data, each equipping a shared base LLM with distinct domain-specific capabilities, activated via self-optimized routing. This allows for dynamic and capability-specific handling of various target tasks, enhancing overall capabilities, without extensive human-labeled data and added parameters. Our empirical results reveal that specializing LLMs may exhibit potential trade-offs in performances on non-specialized tasks. On the other hand, our Self-MoE demonstrates substantial improvements (6.5%p on average) over the base LLM across diverse benchmarks such as knowledge, reasoning, math, and coding. It also consistently outperforms other methods, including instance merging and weight merging, while offering better flexibility and interpretability by design with semantic experts and routing. Our findings highlight the critical role of modularity, the applicability of Self-MoE to multiple base LLMs, and the potential of self-improvement in achieving efficient, scalable, and adaptable systems.

cs.CL↗

arXiv2023

Self-Specialization: Uncovering Latent Expertise within Large Language Models

Recent works have demonstrated the effectiveness of self-alignment in which a large language model is aligned to follow general instructions using instructional data generated from the model itself starting from a handful of human-written seeds. Instead of general alignment, in this work, we focus on self-alignment for expert domain specialization (e.g., biomedicine, finance). As a preliminary, we quantitively show the marginal effect that generic instruction-following training has on downstream expert domains' performance. To remedy this, we propose self-specialization - allowing for effective model specialization while achieving cross-task generalization by leveraging only a few labeled seeds. Self-specialization offers a data- and parameter-efficient way of "carving out" an expert model out of a generalist pre-trained LLM. Exploring a variety of popular open large models as a base for specialization, our experimental results in both biomedical and financial domains show that our self-specialized models outperform their base models by a large margin, and even larger models that are generally instruction-tuned or that have been adapted to the target domain by other means.

cs.CL↗

arXiv2017

Real-Time Local Volt/VAR Control Under External Disturbances with High PV Penetration

Volt/var control (VVC) of smart PV inverter is becoming one of the most popular solutions to address the voltage challenges associated with high PV penetration. This work focuses on the local droop VVC recommended by the grid integration standards IEEE1547, rule21 and addresses their major challenges i.e. appropriate parameters selection under changing conditions, and the control being vulnerable to instability (or voltage oscillations) and significant steady state error (SSE). This is achieved by proposing a two-layer local real-time adaptive VVC that has two major features i.e. a) it is able to ensure both low SSE and control stability simultaneously without compromising either, and b) it dynamically adapts its parameters to ensure good performance in a wide range of external disturbances such as sudden cloud cover, cloud intermittency, and substation voltage changes. A theoretical analysis and convergence proof of the proposed control is also discussed. The proposed control is implementation friendly as it fits well within the integration standard framework and depends only on the local bus information. The performance is compared with the existing droop VVC methods in several scenarios on a large unbalanced 3-phase feeder with detailed secondary side modeling.

math.OC↗

arXiv2017

Assessment of Optimal Flexibility in Ensemble of Frequency Responsive Loads

Potential of electrical loads in providing grid ancillary services is often limited due to the uncertainties associated with the load behavior. A knowledge of the expected uncertainties with a load control program would invariably yield to better informed control policies, opening up the possibility of extracting the maximal load control potential without affecting grid operations. In the context of frequency responsive load control, a probabilistic uncertainty analysis framework is presented to quantify the expected error between the target and actual load response, under uncertainties in the load dynamics. A closed-form expression of an optimal demand flexibility, minimizing the expected error in actual and committed flexibility, is provided. Analytical results are validated through Monte Carlo simulations of ensembles of electric water heaters.

math.OC↗

arXiv2017

Hierarchical Control Framework for Integrated Coordination between DERs and Demand Response

Demand response represents a significant but largely untapped resource that can greatly enhance the flexibility and reliability of power systems. This paper proposes a hierarchical control framework to facilitate the integrated coordination between distributed energy resources and demand response. The proposed framework consists of coordination and device layers. In the coordination layer, various resource aggregations are optimally coordinated in a distributed manner to achieve the system-level objectives. In the device layer, individual resources are controlled in real time to follow the optimal power dispatch signals received from the coordination layer. For practical applications, a method is presented to determine the utility functions of controllable loads by accounting for the real-time load dynamics and the preferences of individual customers. The effectiveness of the proposed framework is validated by detailed simulation studies.

math.OC↗