Himanshu Gupta

Himanshu Gupta, Shreyas Verma, Ujjwala Anantheswaran et al. · gatech

Multi-modal Large Language Models (MLLMs) exhibit impressive problem-solving abilities in various domains, but their visual comprehension and abstract reasoning skills remain under-evaluated. To this end, we present PolyMATH, a challenging benchmark aimed at evaluating the general cognitive reasoning abilities of MLLMs. PolyMATH comprises 5,000 manually collected high-quality images of cognitive textual and visual challenges across 10 distinct categories, including pattern recognition, spatial reasoning, and relative reasoning. We conducted a comprehensive, and quantitative evaluation of 15 MLLMs using four diverse prompting strategies, including Chain-of-Thought and Step-Back. The best scores achieved on PolyMATH are ~41%, ~36%, and ~27%, obtained by Claude-3.5 Sonnet, GPT-4o and Gemini-1.5 Pro respectively - highlighting the logical and visual complexity of these questions. A further fine-grained error analysis reveals that these models struggle to understand spatial relations and perform drawn-out, high-level reasoning. This is further strengthened by our ablation study estimating MLLM performance when given textual descriptions in place of diagrams. As evidenced by ~4% improvement over textual descriptions as opposed to actual images, we discover that models do not truly comprehend visual diagrams and the spatial information therein, and are thus prone to logical errors. Finally, we evaluate the OpenAI o1 models and find that their performance only matches the human baseline, highlighting the difficulty of the benchmark. The results on PolyMATH highlight the room for improvement in multi-modal reasoning and provide unique insights to guide the development of future MLLMs.

Mihir Parmar, Aakanksha Naik, Himanshu Gupta et al.

Many large language models (LLMs) for medicine have largely been evaluated on short texts, and their ability to handle longer sequences such as a complete electronic health record (EHR) has not been systematically explored. Assessing these models on long sequences is crucial since prior work in the general domain has demonstrated performance degradation of LLMs on longer texts. Motivated by this, we introduce LongBoX, a collection of seven medical datasets in text-to-text format, designed to investigate model performance on long sequences. Preliminary experiments reveal that both medical LLMs (e.g., BioGPT) and strong general domain LLMs (e.g., FLAN-T5) struggle on this benchmark. We further evaluate two techniques designed for long-sequence handling: (i) local-global attention, and (ii) Fusion-in-Decoder (FiD). Our results demonstrate mixed results with long-sequence handling - while scores on some datasets increase, there is substantial room for improvement. We hope that LongBoX facilitates the development of more effective long-sequence techniques for the medical domain. Data and source code are available at https://github.com/Mihir3009/LongBoX.

Himanshu Gupta, Kevin Scaria, Ujjwala Anantheswaran et al. · gatech

The rapid advancement of large language models (LLMs) has sparked interest in data synthesis techniques, aiming to generate diverse and high-quality synthetic datasets. However, these synthetic datasets often suffer from a lack of diversity and added noise. In this paper, we present TarGEN, a multi-step prompting strategy for generating high-quality synthetic datasets utilizing a LLM. An advantage of TarGEN is its seedless nature; it does not require specific task instances, broadening its applicability beyond task replication. We augment TarGEN with a method known as self-correction empowering LLMs to rectify inaccurately labeled instances during dataset creation, ensuring reliable labels. To assess our technique's effectiveness, we emulate 8 tasks from the SuperGLUE benchmark and finetune various language models, including encoder-only, encoder-decoder, and decoder-only models on both synthetic and original training sets. Evaluation on the original test set reveals that models trained on datasets generated by TarGEN perform approximately 1-2% points better than those trained on original datasets (82.84% via syn. vs. 81.12% on og. using Flan-T5). When incorporating instruction tuning, the performance increases to 84.54% on synthetic data vs. 81.49% on original data by Flan-T5. A comprehensive analysis of the synthetic dataset compared to the original dataset reveals that the synthetic dataset demonstrates similar or higher levels of dataset complexity and diversity. Furthermore, the synthetic dataset displays a bias level that aligns closely with the original dataset. Finally, when pre-finetuned on our synthetic SuperGLUE dataset, T5-3B yields impressive results on the OpenLLM leaderboard, surpassing the model trained on the Self-Instruct dataset by 4.14% points. We hope that TarGEN can be helpful for quality data generation and reducing the human efforts to create complex benchmarks.

Yunfan Li, Vinayak Shenoy, Prateek Prasanna et al.

Automatic recognition of surgical phases in surgical videos is a fundamental task in surgical workflow analysis. In this report, we propose a Transformer-based method that utilizes calibrated confidence scores for a 2-stage inference pipeline, which dynamically switches between a baseline model and a separately trained transition model depending on the calibrated confidence level. Our method outperforms the baseline model on the Cholec80 dataset, and can be applied to a variety of action segmentation methods.

Pratik Jayarao, Himanshu Gupta, Neeraj Varshney et al. · amazon-science

As Large Language Models (LLMs) are increasingly adopted as automated judges in benchmarking and reward modeling, ensuring their reliability, efficiency, and robustness has become critical. In this work, we present a systematic comparison of "thinking" and "non-thinking" LLMs in the LLM-as-a-judge paradigm using open-source Qwen 3 models of relatively small sizes (0.6B, 1.7B, and 4B parameters). We evaluate both accuracy and computational efficiency (FLOPs) on RewardBench tasks, and further examine augmentation strategies for non-thinking models, including in-context learning, rubric-guided judging, reference-based evaluation, and n-best aggregation. Our results show that despite these enhancements, non-thinking models generally fall short of their thinking counterparts. Our results show that thinking models achieve approximately 10% points higher accuracy with little overhead (under 2x), in contrast to augmentation strategies like few-shot learning, which deliver modest gains at a higher cost (>8x). Bias and robustness analyses further demonstrate that thinking models maintain significantly greater consistency under a variety of bias conditions such as positional, bandwagon, identity, diversity, and random biases (6% higher on average). We further extend our experiments to the multilingual setting and our results confirm that explicit reasoning extends its benefits beyond English. Overall, our work results in several important findings that provide systematic evidence that explicit reasoning offers clear advantages in the LLM-as-a-judge paradigm not only in accuracy and efficiency but also in robustness.

Ujjwala Anantheswaran, Himanshu Gupta, Mihir Parmar et al.

Event detection refers to identifying event occurrences in a text and comprises of two subtasks; event identification and classification. We present EDM3, a novel approach for Event Detection that formulates three generative tasks: identification, classification, and combined detection. We show that EDM3 helps to learn transferable knowledge that can be leveraged to perform Event Detection and its subtasks concurrently, mitigating the error propagation inherent in pipelined approaches. Unlike previous dataset- or domain-specific approaches, EDM3 utilizes the existing knowledge of language models, allowing it to be trained over any classification schema. We evaluate EDM3 on multiple event detection datasets: RAMS, WikiEvents, MAVEN, and MLEE, showing that EDM3 outperforms 1) single-task performance by 8.4% on average and 2) multi-task performance without instructional prompts by 2.4% on average. We obtain SOTA results on RAMS (71.3% vs. 65.1% F-1) and competitive performance on other datasets. We analyze our approach to demonstrate its efficacy in low-resource and multi-sentence settings. We also show the effectiveness of this approach on non-standard event configurations such as multi-word and multi-class event triggers. Overall, our results show that EDM3 is a promising approach for Event Detection that has the potential for real-world applications.

Himanshu Gupta, Shreyas Verma, Santosh Mashetty et al.

Named Entity Recognition (NER) has seen significant progress in recent years, with numerous state-of-the-art (SOTA) models achieving high performance. However, very few studies have focused on the generation of entities' context. In this paper, we introduce CONTEXT-NER, a task that aims to generate the relevant context for entities in a sentence, where the context is a phrase describing the entity but not necessarily present in the sentence. To facilitate research in this task, we also present the EDGAR10-Q dataset, which consists of annual and quarterly reports from the top 1500 publicly traded companies. The dataset is the largest of its kind, containing 1M sentences, 2.8M entities, and an average of 35 tokens per sentence, making it a challenging dataset. We propose a baseline approach that combines a phrase generation algorithm with inferencing using a 220M language model, achieving a ROUGE-L score of 27% on the test split. Additionally, we perform a one-shot inference with ChatGPT, which obtains a 30% ROUGE-L, highlighting the difficulty of the dataset. We also evaluate models such as T5 and BART, which achieve a maximum ROUGE-L of 49% after supervised finetuning on EDGAR10-Q. We also find that T5-large, when pre-finetuned on EDGAR10-Q, achieve SOTA results on downstream finance tasks such as Headline, FPB, and FiQA SA, outperforming vanilla version by 10.81 points. To our surprise, this 66x smaller pre-finetuned model also surpasses the finance-specific LLM BloombergGPT-50B by 15 points. We hope that our dataset and generated artifacts will encourage further research in this direction, leading to the development of more sophisticated language models for financial text analysis

Himanshu Gupta, Abhiram Anand Gulanikar, Lov Kumar et al.

A code smell is a surface indicator of an inherent problem in the system, most often due to deviation from standard coding practices on the developers part during the development phase. Studies observe that code smells made the code more susceptible to call for modifications and corrections than code that did not contain code smells. Restructuring the code at the early stage of development saves the exponentially increasing amount of effort it would require to address the issues stemming from the presence of these code smells. Instead of using traditional features to detect code smells, we use user comments to manually construct features to predict code smells. We use three Extreme learning machine kernels over 629 packages to identify eight code smells by leveraging feature engineering aspects and using sampling techniques. Our findings indicate that the radial basis functional kernel performs best out of the three kernel methods with a mean accuracy of 98.52.

Pradyumna Gupta, Himanshu Gupta, Aman Sinha

Memes have become an ubiquitous social media entity and the processing and analysis of suchmultimodal data is currently an active area of research. This paper presents our work on theMemotion Analysis shared task of SemEval 2020, which involves the sentiment and humoranalysis of memes. We propose a system which uses different bimodal fusion techniques toleverage the inter-modal dependency for sentiment and humor classification tasks. Out of all ourexperiments, the best system improved the baseline with macro F1 scores of 0.357 on SentimentClassification (Task A), 0.510 on Humor Classification (Task B) and 0.312 on Scales of SemanticClasses (Task C).

Himanshu Gupta, Dharanipragada Janakiram

Blockchain is maintained as a global log between a network of nodes and uses cryptographic distributed protocols to synchronize the updates. As adopted by Bitcoin and Ethereum these update operations to the ledger are serialized, and executed in batches. To safeguard the system against the generation of conflicting sets of updates and maintain the consistency of the ledger, the frequency of the updates is controlled, which severely affects the performance of the system. This paper presents Converging Directed Acyclic Graph (CDAG), as a substitute for the chain and DAG structures used in other blockchain protocols. CDAG allows multiple parallel updates to the ledger and converges them at the next step providing finality to the blocks. It partitions the updates into non-intersecting buckets of transactions to prevent the generation of conflicting blocks and divide the time into slots to provide enough time for them to propagate in the network. Multiple simultaneous updates improve the throughput of CDAG, and the converging step helps to finalize them faster, even in the presence of conflicts. Moreover, CDAG provides a total order among the blocks of the ledger to support smart contracts, unlike some of the other blockDAG protocols. We evaluate the performance of CDAG on Google Cloud Platform using Google Kubernetes Engine, simulating a real-time network. Experimental results show that CDAG achieves a throughput of more than 2000 transactions per second and confirms them well in under 2 minutes. Also, the protocol scales well in comparison to other permissioned protocols, and the capacity of the network only limits the performance.