Vivek Srikumar
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
Vivek Srikumar has a total of 117 co-authors. Here are the top co-authors by number of shared publications:
- Martha Palmer
The University of Colorado Boulder
4 shared publications - Nathan Schneider
Georgetown University
4 shared publications - Zac Imel
Department of Educational Psychology, University of Utah
4 shared publications - Dan Roth
University of Pennsylvania
3 shared publications - David C. Atkins
Department of Psychiatry and Behavioral Sciences, University of Washington
3 shared publications Maitrey Mehta
University of Utah
3 shared publications
Latest Publications
- Open Access
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 4298-4304
LACONIC: Dense-Level Effectiveness for Scalable Sparse Retrieval via a Two-Phase Training Curriculum
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 4298-4304
Open AccessLACONIC: Dense-Level Effectiveness for Scalable Sparse Retrieval via a Two-Phase Training Curriculum
- Open Access
LACONIC: Dense-Level Effectiveness for Scalable Sparse Retrieval via a Two-Phase Training Curriculum
While dense retrieval models have become the standard for state-of-the-art information retrieval, their deployment is often constrained by high memory requirements and reliance on GPU accelerators for vector similarity search. Learned sparse retrieval offers a compelling alternative by enabling efficient search via inverted indices, yet it has historically received less attention than dense approaches. In this report, we introduce LACONIC, a family of learned sparse retrievers based on the Llama-3 architecture (1B, 3B, and 8B). We propose a streamlined two-phase training curriculum consisting of (1) weakly supervised pre-finetuning to adapt causal LLMs for bidirectional contextualization and (2) high-signal finetuning using curated hard negatives. Our results demonstrate that LACONIC effectively bridges the performance gap with dense models: the 8B variant achieves a state-of-the-art 60.2 nDCG on the MTEB Retrieval benchmark, ranking 15th on the leaderboard as of January 1, 2026, while utilizing 71\% less index memory than an equivalent dense model. By delivering high retrieval effectiveness on commodity CPU hardware with a fraction of the compute budget required by competing models, LACONIC provides a scalable and efficient solution for real-world search applications.
Open AccessLACONIC: Dense-Level Effectiveness for Scalable Sparse Retrieval via a Two-Phase Training Curriculum
While dense retrieval models have become the standard for state-of-the-art information retrieval, their deployment is often constrained by high memory requirements and reliance on GPU accelerators for vector similarity search. Learned sparse retrieval offers a compelling alternative by enabling efficient search via inverted indices, yet it has historically received less attention than dense approaches. In this report, we introduce LACONIC, a family of learned sparse retrievers based on the Llama-3 architecture (1B, 3B, and 8B). We propose a streamlined two-phase training curriculum consisting of (1) weakly supervised pre-finetuning to adapt causal LLMs for bidirectional contextualization and (2) high-signal finetuning using curated hard negatives. Our results demonstrate that LACONIC effectively bridges the performance gap with dense models: the 8B variant achieves a state-of-the-art 60.2 nDCG on the MTEB Retrieval benchmark, ranking 15th on the leaderboard as of January 1, 2026, while utilizing 71\% less index memory than an equivalent dense model. By delivering high retrieval effectiveness on commodity CPU hardware with a fraction of the compute budget required by competing models, LACONIC provides a scalable and efficient solution for real-world search applications.
- Open Access
- 3 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
CrysText: A Generative AI Approach for Text-Conditioned Crystal Structure Generation using LLM
The ability to generate crystal structures directly from textual descriptions marks a pivotal advancement in materials informatics and underscores the emerging role of large language models (LLMs) in inverse design. In this work, we introduce CrysText, a text-conditioned framework that generates crystal structures in Crystallographic Information File (CIF) format from natural language prompts specifying composition and space group. Leveraging LLaMA-3.1-8B and Mistral-7B-v0.3 fine-tuned using Quantized Low-Rank Adaptation (QLoRA), our approach enables the efficient and scalable generation of CIF-formatted structures directly from input descriptions, eliminating the need for post-processing with rapid inference. Evaluations on the MP-20 benchmark demonstrate high structural match rates and low RMSE values, confirming the model’s ability to generate physically consistent crystal structures aligned with compositional and symmetry constraints. By incorporating energy above the convex hull as a conditioning parameter, CrysText further demonstrates the ability to generate thermodynamically stable novel materials. We subsequently extend this framework with CrysText-RL, which integrates Group Relative Policy Optimization (GRPO) to provide reinforcement learning feedback directly on generated CIF outputs via group-based normalized rewards. CrysText-RL achieves additional improvements over the supervised CrysText model in terms of composition and space group satisfiability and structure match rate. This work establishes a scalable paradigm for text-driven crystal structure generation, demonstrating that both supervised fine-tuning and reinforcement learning enable a pathway towards accelerated materials discovery.
- 3 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessCrysText: A Generative AI Approach for Text-Conditioned Crystal Structure Generation using LLM
The ability to generate crystal structures directly from textual descriptions marks a pivotal advancement in materials informatics and underscores the emerging role of large language models (LLMs) in inverse design. In this work, we introduce CrysText, a text-conditioned framework that generates crystal structures in Crystallographic Information File (CIF) format from natural language prompts specifying composition and space group. Leveraging LLaMA-3.1-8B and Mistral-7B-v0.3 fine-tuned using Quantized Low-Rank Adaptation (QLoRA), our approach enables the efficient and scalable generation of CIF-formatted structures directly from input descriptions, eliminating the need for post-processing with rapid inference. Evaluations on the MP-20 benchmark demonstrate high structural match rates and low RMSE values, confirming the model’s ability to generate physically consistent crystal structures aligned with compositional and symmetry constraints. By incorporating energy above the convex hull as a conditioning parameter, CrysText further demonstrates the ability to generate thermodynamically stable novel materials. We subsequently extend this framework with CrysText-RL, which integrates Group Relative Policy Optimization (GRPO) to provide reinforcement learning feedback directly on generated CIF outputs via group-based normalized rewards. CrysText-RL achieves additional improvements over the supervised CrysText model in terms of composition and space group satisfiability and structure match rate. This work establishes a scalable paradigm for text-driven crystal structure generation, demonstrating that both supervised fine-tuning and reinforcement learning enable a pathway towards accelerated materials discovery.
- 27 September 2025
- Published by Springer Nature
- p. 38-55
The Role of Transformer Architecture in the Logic-as-Loss Framework
- 27 September 2025
- Published by Springer Nature
- p. 38-55
The Role of Transformer Architecture in the Logic-as-Loss Framework
- Open Access
Unequal Voices: How LLMs Construct Constrained Queer Narratives
One way social groups are marginalized in discourse is that the narratives told about them often default to a narrow, stereotyped range of topics. In contrast, default groups are allowed the full complexity of human existence. We describe the constrained representations of queer people in LLM generations in terms of harmful representations, narrow representations, and discursive othering and formulate hypotheses to test for these phenomena. Our results show that LLMs are significantly limited in their portrayals of queer personas.
Open AccessUnequal Voices: How LLMs Construct Constrained Queer Narratives
One way social groups are marginalized in discourse is that the narratives told about them often default to a narrow, stereotyped range of topics. In contrast, default groups are allowed the full complexity of human existence. We describe the constrained representations of queer people in LLM generations in terms of harmful representations, narrow representations, and discursive othering and formulate hypotheses to test for these phenomena. Our results show that LLMs are significantly limited in their portrayals of queer personas.
- Open Access
Reinforcing Code Generation: Improving Text-to-SQL with Execution-Based Learning
In this work, we study the problem of code generation with a large language model (LLM), with a focus on generating SQL queries from natural language questions. We ask: Instead of using supervised fine tuning with text-code pairs, can we tune a model by having it interact with a database engine? We frame this problem as a reinforcement learning problem where the model receives execution-based feedback from the environment in the form of scalar rewards. These rewards penalize execution failures and assign positive values when a query returns a correct answer. We use the rewards within the Group Relative Policy Optimization (GRPO) framework. We use a tabular reasoning benchmark to test and evaluate our findings. We find that with only weak supervision in the form of question-answer pairs, RL-tuning improves the accuracy of model generated SQL code from 31.49 to 49.83 while reducing error percentage from 25.43% to 14.71%. This improvement allowed the model nearly match the performance performance to the larger SQLCoder-70B model. Our work demonstrates the potential of using execution-based feedback to improve symbolic reasoning capabilities of LLMs.
Open AccessReinforcing Code Generation: Improving Text-to-SQL with Execution-Based Learning
In this work, we study the problem of code generation with a large language model (LLM), with a focus on generating SQL queries from natural language questions. We ask: Instead of using supervised fine tuning with text-code pairs, can we tune a model by having it interact with a database engine? We frame this problem as a reinforcement learning problem where the model receives execution-based feedback from the environment in the form of scalar rewards. These rewards penalize execution failures and assign positive values when a query returns a correct answer. We use the rewards within the Group Relative Policy Optimization (GRPO) framework. We use a tabular reasoning benchmark to test and evaluate our findings. We find that with only weak supervision in the form of question-answer pairs, RL-tuning improves the accuracy of model generated SQL code from 31.49 to 49.83 while reducing error percentage from 25.43% to 14.71%. This improvement allowed the model nearly match the performance performance to the larger SQLCoder-70B model. Our work demonstrates the potential of using execution-based feedback to improve symbolic reasoning capabilities of LLMs.
- Open Access
LLM-Symbolic Integration for Robust Temporal Tabular Reasoning
Temporal tabular question answering presents a significant challenge for Large Language Models (LLMs), requiring robust reasoning over structured data, which is a task where traditional prompting methods often fall short. These methods face challenges such as memorization, sensitivity to table size, and reduced performance on complex queries. To overcome these limitations, we introduce TempTabQA-C, a synthetic dataset designed for systematic and controlled evaluations, alongside a symbolic intermediate representation that transforms tables into database schemas. This structured approach allows LLMs to generate and execute SQL queries, enhancing generalization and mitigating biases. By incorporating adaptive few-shot prompting with contextually tailored examples, our method achieves superior robustness, scalability, and performance. Experimental results consistently highlight improvements across key challenges, setting a new benchmark for robust temporal reasoning with LLMs.
Open AccessLLM-Symbolic Integration for Robust Temporal Tabular Reasoning
Temporal tabular question answering presents a significant challenge for Large Language Models (LLMs), requiring robust reasoning over structured data, which is a task where traditional prompting methods often fall short. These methods face challenges such as memorization, sensitivity to table size, and reduced performance on complex queries. To overcome these limitations, we introduce TempTabQA-C, a synthetic dataset designed for systematic and controlled evaluations, alongside a symbolic intermediate representation that transforms tables into database schemas. This structured approach allows LLMs to generate and execute SQL queries, enhancing generalization and mitigating biases. By incorporating adaptive few-shot prompting with contextually tailored examples, our method achieves superior robustness, scalability, and performance. Experimental results consistently highlight improvements across key challenges, setting a new benchmark for robust temporal reasoning with LLMs.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 19914-19940
LLM-Symbolic Integration for Robust Temporal Tabular Reasoning
Atharv Kulkarni, Kushagra Dixit, Vivek Srikumar, Dan Roth, Vivek Gupta. Findings of the Association for Computational Linguistics: ACL 2025. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 19914-19940
Open AccessLLM-Symbolic Integration for Robust Temporal Tabular Reasoning
Atharv Kulkarni, Kushagra Dixit, Vivek Srikumar, Dan Roth, Vivek Gupta. Findings of the Association for Computational Linguistics: ACL 2025. 2025.
- Open Access
- 18 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 64-74
LogFlux: A Software Suite for Replicating Results in Automated Log Parsing
Logging mechanisms are a cornerstone in the development and maintenance of computer systems, gaining even greater prominence in the era of large-scale cloud-based applications. Their critical role in real-time system monitoring and behavior analysis cannot be overstated, making them invaluable tools for system administrators and developers alike. Automated parsing of log messages—turning the text of log messages into parsed, structured data—is a significant research area. Despite the number of log parsers that have emerged over the years, there has been a noticeable gap in the evaluation of these tools, reproduction of results, and direct comparisons between them on a level playing field. Recognizing this, we re-implemented twelve of the most popular log parsers from scratch, enabling them to be used in replication studies. This paper presents our open-source project LogFlux, which is a suite for evaluating automated log parsers so that studies involving them can be replicated. Through LogFlux, we aim to bridge the gap between theoretical log parsing methods and their practical application, offering a robust and easy to use solution that is accessible and effective for a range of users. Our experience in attempting to obtain results from many published log parser algorithms has shed light on important aspects of replication, such as the value of independent implementation for uncovering bugs and the need for careful software engineering to facilitate maintenance.
- 18 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 64-74
Open AccessLogFlux: A Software Suite for Replicating Results in Automated Log Parsing
Logging mechanisms are a cornerstone in the development and maintenance of computer systems, gaining even greater prominence in the era of large-scale cloud-based applications. Their critical role in real-time system monitoring and behavior analysis cannot be overstated, making them invaluable tools for system administrators and developers alike. Automated parsing of log messages—turning the text of log messages into parsed, structured data—is a significant research area. Despite the number of log parsers that have emerged over the years, there has been a noticeable gap in the evaluation of these tools, reproduction of results, and direct comparisons between them on a level playing field. Recognizing this, we re-implemented twelve of the most popular log parsers from scratch, enabling them to be used in replication studies. This paper presents our open-source project LogFlux, which is a suite for evaluating automated log parsers so that studies involving them can be replicated. Through LogFlux, we aim to bridge the gap between theoretical log parsing methods and their practical application, offering a robust and easy to use solution that is accessible and effective for a range of users. Our experience in attempting to obtain results from many published log parser algorithms has shed light on important aspects of replication, such as the value of independent implementation for uncovering bugs and the need for careful software engineering to facilitate maintenance.
- Open Access
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 1407-1418
An In-depth Investigation of User Response Simulation for Conversational Search
Conversational search has seen increased recent attention in both the IR and NLP communities. It seeks to clarify and solve users' search needs through multi-turn natural language interactions. However, most existing systems are trained and demonstrated with recorded or artificial conversation logs. Eventually, conversational search systems should be trained, evaluated, and deployed in an open-ended setting with unseen conversation trajectories. A key challenge is that training and evaluating such systems both require a human-in-the-loop, which is expensive and does not scale. One strategy is to simulate users, thereby reducing the scaling costs. However, current user simulators are either limited to only responding to yes-no questions from the conversational search system or unable to produce high-quality responses in general. This paper shows that existing user simulation systems could be significantly improved by a smaller finetuned natural language generation model. However, rather than merely reporting it as the new state-of-the-art, we consider it a strong baseline and present an in-depth investigation of simulating user response for conversational search. Our goal is to supplement existing work with an insightful hand-analysis of unsolved challenges by the baseline and propose our solutions. The challenges we identified include (1) a blind spot that is difficult to learn, and (2) a specific type of misevaluation in the standard setup. We propose a new generation system to effectively cover the training blind spot and suggest a new evaluation setup to avoid misevaluation. Our proposed system leads to significant improvements over existing systems and large language models such as GPT-4. Additionally, our analysis provides insights into the nature of the task to facilitate future work.
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 1407-1418
Open AccessAn In-depth Investigation of User Response Simulation for Conversational Search
Conversational search has seen increased recent attention in both the IR and NLP communities. It seeks to clarify and solve users' search needs through multi-turn natural language interactions. However, most existing systems are trained and demonstrated with recorded or artificial conversation logs. Eventually, conversational search systems should be trained, evaluated, and deployed in an open-ended setting with unseen conversation trajectories. A key challenge is that training and evaluating such systems both require a human-in-the-loop, which is expensive and does not scale. One strategy is to simulate users, thereby reducing the scaling costs. However, current user simulators are either limited to only responding to yes-no questions from the conversational search system or unable to produce high-quality responses in general. This paper shows that existing user simulation systems could be significantly improved by a smaller finetuned natural language generation model. However, rather than merely reporting it as the new state-of-the-art, we consider it a strong baseline and present an in-depth investigation of simulating user response for conversational search. Our goal is to supplement existing work with an insightful hand-analysis of unsolved challenges by the baseline and propose our solutions. The challenges we identified include (1) a blind spot that is difficult to learn, and (2) a specific type of misevaluation in the standard setup. We propose a new generation system to effectively cover the training blind spot and suggest a new evaluation setup to avoid misevaluation. Our proposed system leads to significant improvements over existing systems and large language models such as GPT-4. Additionally, our analysis provides insights into the nature of the task to facilitate future work.