Sangharatna Godboley
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sangharatna Godboley has a total of 65 co-authors. Here are the top co-authors by number of shared publications:
Durga Prasad Mohapatra
12 shared publicationsArpita Dutta
Indian Institute of Technology Kharagpur
11 shared publications- Durga Prasad Mohapatra
Department of Computer Science and Engineering, National Institute of Technology Rourkela, Rourkela, India
10 shared publications - Radha Krishna Pisipati
National Institute of Technology Warangal
8 shared publications - Arpita Dutta
National University of Singapore
7 shared publications - Monika Rani Golla
Department of Computer Science and Engineering, National Institute of Technology Warangal, Warangal, India
6 shared publications
Latest Publications
- Open Access
- 10 October 2026
- Published by Elsevier in SSRN Electronic Journal
When Does an Offline LLM Help a Learned SDC Risk Model? A Held-Out Study of Risk-Semantic Fusion
Running every generated SDC test scenario is too costly, so scenarios must be ranked to prioritize likely failures early. KSERESNET is a learned risk-ranking model; this study investigates whether an offline large language model (LLM) provides complementary ranking information for KSERESNET’s poorly-ranked failures. We investigate whether fusion gains observed with the frozen checkpoint can be attributed specifically to the LLM, and whether such gains persist once KSERESNET is retrained. We evaluate Qwen3-8B, combined with the frozen checkpoint via fixed convex fusion and a learned Adaptive Risk-Semantic Fusion (ARSF) gate. We separately retrain KSERESNET with data-derived normalization, score scenarios with GBDT and an independent Transformer, and evaluate multi-way fusion against this retrained baseline, with tuning performed on validation data. Against the frozen checkpoint (APFD 0.6334), fixed fusion and ARSF both improve ranking, ARSF giving the strongest held-out performance (APFD 0.6534). Retraining KSERESNET alone reaches APFD 0.7577, comparable to an independent Transformer (0.7535). Against this retrained baseline, grid-searched multi-way fusion finds no statistically detectable improvement from adding the LLM, with the search itself assigning it negligible weight: the complementarity apparent against the frozen checkpoint does not survive baseline correction. This is itself the paper’s central finding: the frozen checkpoint is insufficient for attributing fusion gains to the LLM, apparent LLM-fusion benefits depend on baseline strength, and should not be generalized as evidence that offline LLMs improve properly trained risk models. Baseline strength should be verified before attributing fusion gains to an added signal.
- 10 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessWhen Does an Offline LLM Help a Learned SDC Risk Model? A Held-Out Study of Risk-Semantic Fusion
Running every generated SDC test scenario is too costly, so scenarios must be ranked to prioritize likely failures early. KSERESNET is a learned risk-ranking model; this study investigates whether an offline large language model (LLM) provides complementary ranking information for KSERESNET’s poorly-ranked failures. We investigate whether fusion gains observed with the frozen checkpoint can be attributed specifically to the LLM, and whether such gains persist once KSERESNET is retrained. We evaluate Qwen3-8B, combined with the frozen checkpoint via fixed convex fusion and a learned Adaptive Risk-Semantic Fusion (ARSF) gate. We separately retrain KSERESNET with data-derived normalization, score scenarios with GBDT and an independent Transformer, and evaluate multi-way fusion against this retrained baseline, with tuning performed on validation data. Against the frozen checkpoint (APFD 0.6334), fixed fusion and ARSF both improve ranking, ARSF giving the strongest held-out performance (APFD 0.6534). Retraining KSERESNET alone reaches APFD 0.7577, comparable to an independent Transformer (0.7535). Against this retrained baseline, grid-searched multi-way fusion finds no statistically detectable improvement from adding the LLM, with the search itself assigning it negligible weight: the complementarity apparent against the frozen checkpoint does not survive baseline correction. This is itself the paper’s central finding: the frozen checkpoint is insufficient for attributing fusion gains to the LLM, apparent LLM-fusion benefits depend on baseline strength, and should not be generalized as evidence that offline LLMs improve properly trained risk models. Baseline strength should be verified before attributing fusion gains to an added signal.
- 23 June 2026
- Published by Springer Nature
- p. 64-73
Heterogeneous Transfer Learning Framework for Semi-supervised Intrusion Detection Across Network Domains
- 23 June 2026
- Published by Springer Nature
- p. 64-73
Heterogeneous Transfer Learning Framework for Semi-supervised Intrusion Detection Across Network Domains
- Open Access
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
A Data-Driven Framework for Evaluating Mobile UI Usability through Interaction Elements and Design Semantics
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Open AccessA Data-Driven Framework for Evaluating Mobile UI Usability through Interaction Elements and Design Semantics
- Open Access
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Workshop Report: From Testing Automation to Fault Prediction using LLM at ISEC 2026
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Open AccessWorkshop Report: From Testing Automation to Fault Prediction using LLM at ISEC 2026
- Open Access
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
kS-LLM: k-Step based Automatic LLM Test Case Generator using Caching Mechanism to Achieve Higher Code Coverage
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Open AccesskS-LLM: k-Step based Automatic LLM Test Case Generator using Caching Mechanism to Achieve Higher Code Coverage
- Open Access
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-4
Towards Reliable Task-Based Chatbots: Coverage, Mutation, and Repair
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-4
Open AccessTowards Reliable Task-Based Chatbots: Coverage, Mutation, and Repair
- Open Access
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-10
cfgLLM: Automated assistance for software requirements using multi-LLM Fusion Agentic Framework
A software requirements analysis document comprises various functional and non-functional requirements of stakeholders and is often expressed in structured natural language. If these requirements are ambiguous or deviate from the end goals, it leads to significant delays and inefficiencies during the software development process. Traditionally, control flow graphs (CFGs) are used to represent the logical flow and complexity of a software system. However, they are generated from the source code and require human effort and experience. In this paper, we propose a novel multi-LLM fusion agentic framework for automated CFG generation directly from natural language software requirements. This serves as a supporting tool, enabling early identification of ambiguities, inconsistencies, and structural issues before implementation. The framework leverages multiple large language models (ChatGPT, LLaMA 3, and Gemini) to extract and interpret control logic. It employs an additional ChatGPT agent to fuse outputs along with contextual information from a vector database via a retrieved augmented generation (RAG) mechanism. To ensure semantic validation, we proposed evaluation metrics: (i) function accuracy, (ii) event accuracy, and (iii) event consistency. Cyclomatic complexity is also used to assess a program’s structural soundness. Experimental results demonstrate that cfgLLM generated valid CFGs, offering assistance for early-stage software design and requirement validation.
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-10
Open AccesscfgLLM: Automated assistance for software requirements using multi-LLM Fusion Agentic Framework
A software requirements analysis document comprises various functional and non-functional requirements of stakeholders and is often expressed in structured natural language. If these requirements are ambiguous or deviate from the end goals, it leads to significant delays and inefficiencies during the software development process. Traditionally, control flow graphs (CFGs) are used to represent the logical flow and complexity of a software system. However, they are generated from the source code and require human effort and experience. In this paper, we propose a novel multi-LLM fusion agentic framework for automated CFG generation directly from natural language software requirements. This serves as a supporting tool, enabling early identification of ambiguities, inconsistencies, and structural issues before implementation. The framework leverages multiple large language models (ChatGPT, LLaMA 3, and Gemini) to extract and interpret control logic. It employs an additional ChatGPT agent to fuse outputs along with contextual information from a vector database via a retrieved augmented generation (RAG) mechanism. To ensure semantic validation, we proposed evaluation metrics: (i) function accuracy, (ii) event accuracy, and (iii) event consistency. Cyclomatic complexity is also used to assess a program’s structural soundness. Experimental results demonstrate that cfgLLM generated valid CFGs, offering assistance for early-stage software design and requirement validation.
- Open Access
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-4
An Integrated Approach to Hybrid Fuzz–Combinatorial and Structured Testcase Generation for Improved Coverage and Fault Detection Efficiency
Modern software test case generation requires techniques that balance structural coverage and input variation to ensure high reliability and robustness. Although boundary-based and combinatorial approaches offer thorough coverage, they might miss unanticipated defects. Conventional fuzzing methods lack systematic coverage but are effective at exposing hidden vulnerabilities through random inputs. This work presents an integrated hybrid framework that integrates fuzzing, combinatorial, and structured testing techniques to increase overall test effectiveness. The proposed approach combines FuzzCombTest, which combines American Fuzzy Lop (AFL) based fuzzing with N-way combinatorial generation using Pairwise Independent Combinatorial Testing (PICT) tool for wider input exploration, with Carbon-Box Testing. This semi-structured technique combines Random, Dictionary, and Random-Dictionary test generation, supported by pairwise analysis to improve branch and line coverage. The framework also incorporates Boundary Value Analysis (BVA), Equivalence Class (EQ), and combined BVA–EQ testing using PICT to guarantee methodical boundary-focused evaluation. Experimental studies on benchmark programs demonstrate significant improvements in coverage precision and execution efficiency, highlighting the potential of this hybrid approach as a scalable and flexible solution for automated software testing.
- 19 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-4
Open AccessAn Integrated Approach to Hybrid Fuzz–Combinatorial and Structured Testcase Generation for Improved Coverage and Fault Detection Efficiency
Modern software test case generation requires techniques that balance structural coverage and input variation to ensure high reliability and robustness. Although boundary-based and combinatorial approaches offer thorough coverage, they might miss unanticipated defects. Conventional fuzzing methods lack systematic coverage but are effective at exposing hidden vulnerabilities through random inputs. This work presents an integrated hybrid framework that integrates fuzzing, combinatorial, and structured testing techniques to increase overall test effectiveness. The proposed approach combines FuzzCombTest, which combines American Fuzzy Lop (AFL) based fuzzing with N-way combinatorial generation using Pairwise Independent Combinatorial Testing (PICT) tool for wider input exploration, with Carbon-Box Testing. This semi-structured technique combines Random, Dictionary, and Random-Dictionary test generation, supported by pairwise analysis to improve branch and line coverage. The framework also incorporates Boundary Value Analysis (BVA), Equivalence Class (EQ), and combined BVA–EQ testing using PICT to guarantee methodical boundary-focused evaluation. Experimental studies on benchmark programs demonstrate significant improvements in coverage precision and execution efficiency, highlighting the potential of this hybrid approach as a scalable and flexible solution for automated software testing.
- Open Access
- 17 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 840-845
Validation Framework for E-Contract and Smart Contract
- 17 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 840-845
Open AccessValidation Framework for E-Contract and Smart Contract
- 11 June 2025
- Published by Elsevier in Journal of Systems and Software
- Vol. 230
ROR-DSE: ROR adequate test case generation using dynamic symbolic execution
- 11 June 2025
- Published by Elsevier in Journal of Systems and Software
- Vol. 230
ROR-DSE: ROR adequate test case generation using dynamic symbolic execution