arXiv subjects

Alexander Tessier

Publications and source records attributed to Alexander Tessier.

6 recordsLinked to original sources

arXiv2026

Rethinking Privileged Information in On-Policy Self-Distillation

On-policy self-distillation (OPSD) trains a student on its own responses using token-level supervision from the same model conditioned on privileged reference information. We investigate whether performance gains from OPSD show that the student learned the information in the reference or instead reflect recovery of reasoning behavior already present in the base model. We perform OPSD experiments on science and mathematics datasets using Qwen3 models ranging from 1.7B to 8B. Our analysis framework separates the supervision induced by the reference from the supervision provided by the teacher without the reference and measures how each aligns with changes in the student's predictions. The correct reference does not provide a consistent performance benefit across teacher generation modes, model sizes, and training datasets. Students can improve without the correct reference, and a solution from another problem can outperform the correct solution on several mathematical reasoning benchmarks. The student's predictions align more strongly with the base model's thinking behavior than with the supervision induced by the reference, but controls constructed from other problems reproduce much of both alignments. Moreover, stronger alignment attributable to the correct reference does not reliably coincide with a greater performance benefit from the reference. Performance gains and distributional alignment alone therefore cannot determine how privileged reference information contributes to student learning in OPSD.

cs.LG↗

arXiv2026

CiteCheck: Retrieval-Grounded Detection of LLM Citation Hallucinations in Scientific Text

Large language models (LLMs) are increasingly used to generate scientific reports, but they can produce references that appear plausible while containing corrupted metadata or pointing to papers that do not exist. We introduce CiteCheck, a hybrid framework for citation hallucination detection that verifies whether a citation corresponds to a real scholarly work and whether its metadata is faithful to that work. CiteCheck retrieves candidate publications from external scholarly sources, compares the citation against the retrieved candidate using a structured LLM verifier, and maps verifier scores into three labels: Exact, Minor, and Major. We also construct a 982-citation physics benchmark with controlled corruptions that capture both subtle metadata drift and fully fabricated references. On the held-out test set, CiteCheck achieves 88.7 macro-F1 and 88.9% accuracy, outperforming GPT, Claude, and Gemini baselines, including web-search and few-shot variants. These results show that reliable citation verification benefits from combining scholarly retrieval, structured LLM-based comparison, and calibrated decision rules.

cs.DL↗

arXiv2026

DeepSciVerify: Verifying Scientific Claim--Citation Alignment via LLM-Driven Evidence Escalation

Misalignment between claims and their cited evidence is a common failure mode in reports generated by large language models, limiting their reliability in scientific and other high-stakes settings. We present DeepSciVerify, a two-stage pipeline for scientific claim-citation verification that combines abstract-level reasoning with selective escalation to passage-level evidence. The system first verifies claims using the abstract and defers uncertain cases, retrieving and analyzing full-text passages only when necessary. This design leverages complementary behaviors across LLMs, as some models are more conservative while others are more decisive under uncertainty. On the SCitance benchmark, DeepSciVerify achieves 86.7 Micro-F1, outperforming strong abstract-only baselines by +4.5 points while resolving 67% of instances without full-text retrieval. These results suggest that selective evidence escalation improves both accuracy and efficiency in claim-citation verification.

cs.AI↗

arXiv2025

Transformer-Based Interfaces for Mechanical Assembly Design: A Gear Train Case Study

Generative artificial intelligence (AI), particularly transformer-based models, presents new opportunities for automating and augmenting engineering design workflows. However, effectively integrating these models into interactive tools requires careful interface design that leverages their unique capabilities. This paper introduces a transformer model tailored for gear train assembly design, paired with two novel interaction modes: Explore and Copilot. Explore Mode uses probabilistic sampling to generate and evaluate diverse design alternatives, while Copilot Mode utilizes autoregressive prediction to support iterative, context-aware refinement. These modes emphasize key transformer properties (sequence-based generation and probabilistic exploration) to facilitate intuitive and efficient human-AI collaboration. Through a case study, we demonstrate how well-designed interfaces can enhance engineers' ability to balance automation with domain expertise. A user study shows that Explore Mode supports rapid exploration and problem redefinition, while Copilot Mode provides greater control and fosters deeper engagement. Our results suggest that hybrid workflows combining both modes can effectively support complex, creative engineering design processes.

cs.HC↗

arXiv2024

Elicitron: An LLM Agent-Based Simulation Framework for Design Requirements Elicitation

Requirements elicitation, a critical, yet time-consuming and challenging step in product development, often fails to capture the full spectrum of user needs. This may lead to products that fall short of expectations. This paper introduces a novel framework that leverages Large Language Models (LLMs) to automate and enhance the requirements elicitation process. LLMs are used to generate a vast array of simulated users (LLM agents), enabling the exploration of a much broader range of user needs and unforeseen use cases. These agents engage in product experience scenarios, through explaining their actions, observations, and challenges. Subsequent agent interviews and analysis uncover valuable user needs, including latent ones. We validate our framework with three experiments. First, we explore different methodologies for diverse agent generation, discussing their advantages and shortcomings. We measure the diversity of identified user needs and demonstrate that context-aware agent generation leads to greater diversity. Second, we show how our framework effectively mimics empathic lead user interviews, identifying a greater number of latent needs than conventional human interviews. Third, we showcase that LLMs can be used to analyze interviews, capture needs, and classify them as latent or not. Our work highlights the potential of using LLM agents to accelerate early-stage product development, reduce costs, and increase innovation.

cs.HC↗

arXiv2023

Multi-split configuration design for fluid-based thermal management systems

High power density systems require efficient cooling to maintain their thermal performance. Despite this, as systems get larger and more complex, human practice and insight may not suffice to determine the desired thermal management system designs. To this end, a framework for automatic architecture exploration is presented in this article for a class of single-phase, multi-split cooling systems. For this class of systems, heat generation devices are clustered based on their spatial information, and flow-split are added only when required and at the location of heat devices. To generate different architectures, candidate architectures are represented as graphs. From these graphs, dynamic physics models are created automatically using a graph-based thermal modeling framework. Then, an optimal fluid flow distribution problem is solved by addressing temperature constraints in the presence of exogenous heat loads to achieve optimal performance. The focus in this work is on the design of general multi-split heat management systems. The architectures discussed here can be used for various applications in the domain of configuration design. The multi-split algorithm can produce configurations where splitting can occur at any of the vertices. The results presented include 3 categories of cases and are discussed in detail.

eess.SY↗