arXiv subjects

Md Touhidul Islam

Publications and source records attributed to Md Touhidul Islam.

9 recordsLinked to original sources

arXiv2026

CircuitFormer: A Circuit Language Model for Analog Topology Design from Natural Language Prompt

Automating analog circuit design remains a longstanding challenge in Electronic Design Automation (EDA). While Transformer-based Large Language Models (LLMs) have revolutionized software code generation, their application to analog hardware design is hindered by two critical limitations: (i) the scarcity of analog design datasets containing natural language description of a design and its corresponding netlist, and (ii) the inefficiency of general-purpose tokenizers (e.g., Byte Pair Encoding (BPE)) in capturing the inherent graph structure of circuits. To bridge this gap, first, we curate the largest annotated dataset of analog circuit netlists to date, comprising 31,341 netlist-natural language description pairs across all major circuit classes. Furthermore, we propose Circuit Tokenizer (CKT), a novel circuit graph tokenizer designed to encode netlist connectivity by explicitly mining frequent subcircuits. In terms of scalability, CKT overcomes the bottleneck of prior circuit graph serialization methods where vocabulary size scales linearly with maximum number of components in the dataset, n_max, (O(n_max)); instead, CKT decouples vocabulary growth from circuit complexity, achieving a constant O(1) complexity. Empirically, CKT outperforms standard BPE on circuit topology representation, reducing sequence length by 57% and achieving a 2.3x superior compression ratio using a compact, fixed vocabulary of size 512. Leveraging this optimized tokenization, we train a circuit-specific language model, CircuitFormer, a 511M parameter encoder-decoder transformer. Our model achieves 100% syntactic correctness and an 83% functional success rate across all major analog circuit categories, outperforming state-of-the-art open-source LLMs by 10% and 14%, respectively, while requiring 240x fewer parameters. The dataset is publicly available at https://huggingface.co/datasets/touhid314/cktformer-dataset.

cs.AI↗

arXiv2026

ChartZero: Synthetic Priors Enable Zero Shot Chart Data Extraction

Automated data extraction from line charts remains fundamentally bottlenecked by extreme stylistic diversity and a severe scarcity of comprehensively annotated, real-world datasets. Current end-to-end pipelines depend heavily on costly manual annotations, crippling their ability to generalize across arbitrary aesthetics and grid layouts. Furthermore, existing models suffer from two critical failure modes during reconstruction. First, extracting thin, intersecting curves frequently causes structural fragmentation and the erasure of fine visual details, as standard architectures struggle against complex backgrounds. Second, semantic association is notoriously error-prone; current pipelines rely on rigid spatial heuristics that easily break down against the unpredictable legend placements of in-the-wild charts. Finally, measuring true progress is hindered by evaluation protocols that assess isolated sub-tasks rather than holistic, end-to-end data reconstruction. To address these foundational issues, we introduce ChartZero, a parsing framework that leverages synthetic priors to enable robust zero-shot chart data extraction. By training exclusively on a purely synthetic dataset of simple mathematical functions, our model completely bypasses the real-world annotation bottleneck. We overcome curve fragmentation via a novel Global Orthogonal Instance (GOI) loss, and replace brittle spatial rules with an open-vocabulary, Vision-Language Model (VLM)-guided legend matching strategy. Accompanied by a new metric and benchmark specifically designed for full end-to-end reconstruction, our evaluations demonstrate that ChartZero significantly advances generalized plot digitization without requiring real-world supervision. Code and dataset will be released upon acceptance.

cs.CV↗

arXiv2026

Shaping Credibility Judgments in Human-GenAI Partnership via Weaker LLMs: A Transactive Memory Perspective on AI Literacy

Generative AI (GenAI) is increasingly used as a knowledge partner in higher education, raising the need for instructional designs that emphasize AI literacy practices such as evaluating output credibility and maintaining human accountability. Existing AI literacy frameworks focus more on what learners should do than on how these practices are enacted in routine student-GenAI collaboration. We address this gap by framing student-GenAI interaction as a transactive memory partnership, where credibility regulates reliance and verification. To make this process visible during coursework, we used a weaker large language model (LLM): small enough to run on most students' computers during class, helpful enough to support learning, but not so capable that it removes the need for verification. In an undergraduate STEM course, students were randomly assigned to one of three conditions across repeated activities: reflection-first (think first, then consult AI), verification-required (use AI, then evaluate the output), or control (unrestricted use). Students completed a transactive memory survey at three time points (N = 42). Weighted credibility diverged by condition over time. ANCOVA controlling for baseline credibility showed a condition effect at mid-semester, F(2, 38) = 4.02, p = .026, partial eta squared = .175, and a stronger effect at post-intervention, F(2, 38) = 5.48, p = .008, partial eta squared = .224; adjusted means were lowest in reflection-first, intermediate in verification-required, and highest in control. Parallel analyses of specialization and coordination were not significant. These findings suggest that workflow sequencing, deliberate use of weaker LLMs, and accountability cues embedded in assignment instructions can recalibrate students' credibility judgments in GenAI use, with reflection-first producing the strongest downward shift in reliance.

cs.HC↗

arXiv2026

A Dataset for Crucial Object Recognition in Blind and Low-Vision Individuals' Navigation

This paper introduces a dataset for improving real-time object recognition systems to aid blind and low-vision (BLV) individuals in navigation tasks. The dataset comprises 21 videos of BLV individuals navigating outdoor spaces, and a taxonomy of 90 objects crucial for BLV navigation, refined through a focus group study. We also provide object labeling for the 90 objects across 31 video segments created from the 21 videos. A deeper analysis reveals that most contemporary datasets used in training computer vision models contain only a small subset of the taxonomy in our dataset. Preliminary evaluation of state-of-the-art computer vision models on our dataset highlights shortcomings in accurately detecting key objects relevant to BLV navigation, emphasizing the need for specialized datasets. We make our dataset publicly available, offering valuable resources for developing more inclusive navigation systems for BLV individuals.

cs.CV↗

arXiv2025

IKIWISI: An Interactive Visual Pattern Generator for Evaluating the Reliability of Vision-Language Models Without Ground Truth

We present IKIWISI ("I Know It When I See It"), an interactive visual pattern generator for assessing vision-language models in video object recognition when ground truth is unavailable. IKIWISI transforms model outputs into a binary heatmap where green cells indicate object presence and red cells indicate object absence. This visualization leverages humans' innate pattern recognition abilities to evaluate model reliability. IKIWISI introduces "spy objects": adversarial instances users know are absent, to discern models hallucinating on nonexistent items. The tool functions as a cognitive audit mechanism, surfacing mismatches between human and machine perception by visualizing where models diverge from human understanding. Our study with 15 participants found that users considered IKIWISI easy to use, made assessments that correlated with objective metrics when available, and reached informed conclusions by examining only a small fraction of heatmap cells. This approach not only complements traditional evaluation methods through visual assessment of model behavior with custom object sets, but also reveals opportunities for improving alignment between human perception and machine understanding in vision-language systems.

cs.CV↗

arXiv2024

Wheeler: A Three-Wheeled Input Device for Usable, Efficient, and Versatile Non-Visual Interaction

Blind users rely on keyboards and assistive technologies like screen readers to interact with user interface (UI) elements. In modern applications with complex UI hierarchies, navigating to different UI elements poses a significant accessibility challenge. Users must listen to screen reader audio descriptions and press relevant keyboard keys one at a time. This paper introduces Wheeler, a novel three-wheeled, mouse-shaped stationary input device, to address this issue. Informed by participatory sessions, Wheeler enables blind users to navigate up to three hierarchical levels in an app independently using three wheels instead of navigating just one level at a time using a keyboard. The three wheels also offer versatility, allowing users to repurpose them for other tasks, such as 2D cursor manipulation. A study with 12 blind users indicates a significant reduction (40%) in navigation time compared to using a keyboard. Further, a diary study with our blind co-author highlights Wheeler's additional benefits, such as accessing UI elements with partial metadata and facilitating mixed-ability collaboration.

cs.HC↗

arXiv2024

Identifying Crucial Objects in Blind and Low-Vision Individuals' Navigation

This paper presents a curated list of 90 objects essential for the navigation of blind and low-vision (BLV) individuals, encompassing road, sidewalk, and indoor environments. We develop the initial list by analyzing 21 publicly available videos featuring BLV individuals navigating various settings. Then, we refine the list through feedback from a focus group study involving blind, low-vision, and sighted companions of BLV individuals. A subsequent analysis reveals that most contemporary datasets used to train recent computer vision models contain only a small subset of the objects in our proposed list. Furthermore, we provide detailed object labeling for these 90 objects across 31 video segments derived from the original 21 videos. Finally, we make the object list, the 21 videos, and object labeling in the 31 video segments publicly available. This paper aims to fill the existing gap and foster the development of more inclusive and effective navigation aids for the BLV community.

cs.CV↗

arXiv2024

Demonstration of Wheeler: A Three-Wheeled Input Device for Usable, Efficient, and Versatile Non-Visual Interaction

Navigating multi-level menus with complex hierarchies remains a big challenge for blind and low-vision users, who predominantly use screen readers to interact with computers. To that end, we demonstrate Wheeler, a three-wheeled input device with two side buttons that can speed up complex multi-level hierarchy navigation in common applications. When in operation, the three wheels of Wheeler are each mapped to a different level in the application hierarchy. Each level can be independently traversed using its designated wheel, allowing users to navigate through multiple levels efficiently. Wheeler's three wheels can also be repurposed for other tasks such as 2D cursor manipulation. In this demonstration, we describe the different operation modes and usage of Wheeler.

cs.HC↗

arXiv2024

A Temporal Psycholinguistics Approach to Identity Resolution of Social Media Users

In this thesis, we propose an approach to identity resolution across social media platforms using the topics, sentiments, and timings of the posts on the platforms. After collecting the public posts of around 5000 profiles from Disqus and Twitter, we analyze their posts to match their profiles across the two platforms. We pursue both temporal and non-temporal methods in our analysis. While neither approach proves definitively superior, the temporal approach generally performs better. We found that the temporal window size influences results more than the shifting amount. On the other hand, our sentiment analysis shows that the inclusion of sentiment makes little difference, probably due to flawed data extraction methods. We also experimented with a distance-based reward-and-punishment-focused scoring model, which achieved an accuracy of 24.198% and an average rank of 158.217 out of 2525 in our collected corpus. Future work includes refining sentiment analysis by evaluating sentiments per topic, extending temporal analysis with additional phases, and improving the scoring model through weight adjustments and modified rewards.

cs.CL↗