Papers by Ayush Singh
ARISE: Iterative Rule Induction and Synthetic Data Generation for Text Classification (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing frameworks for large language models (LLMs) generate high-quality synthetic data that can be used to supplement training data or surpass crowd-sourced annotations. |
| Approach: | They propose a framework that iteratively induces rules and generates synthetic data for text classification. |
| Outcome: | The proposed framework outperforms existing models on in-context learning and fine-tuning settings by using augmented data. |
The Bull and the Bear: Summarizing Stock Market Discussions (2022.lrec-1)
Copied to clipboard
| Challenge: | a dataset of 7888 reddit posts and 400 posts is used to summarize stock market topics. |
| Approach: | They curate discussions on social media platforms and construct an abstractive summarization dataset. |
| Outcome: | The proposed dataset consists of 7888 Reddit posts and summaries for 400 posts . it is robustly evaluated and will be made publicly available . |
Enhancing Adverse Drug Event Detection with Multimodal Dataset: Corpus Creation and Model Development (2024.findings-acl)
Copied to clipboard
| Challenge: | ADEs are a serious public health concern and cost healthcare systems billions of dollars . despite advancements in healthcare, ADE detection remains a significant challenge . |
| Approach: | They propose a multimodal adverse drug event detection dataset that merges ADE-related textual information with visual aids to enhance patient safety. |
| Outcome: | The proposed dataset integrates ADE-related textual information with visual aids to improve patient safety and healthcare accessibility. |
A Benchmark and Dataset for Post-OCR text correction in Sanskrit (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Sanskrit is a classical language with 30 million manuscripts available for digitisation . however, it is considered to be low-resource when it comes to available digital resources. |
| Approach: | They propose to use a post-OCR text correction dataset to correct errors from OCR predictions from 30 different books in the Indian subcontinent. |
| Outcome: | The proposed model outperforms OCR models on graphemic and lexical levels and shows that it is more accurate than previous models. |
IPO: Your Language Model is Secretly a Preference Classifier (2025.acl-long)
Copied to clipboard
| Challenge: | Reinforcement learning from human feedback (RLHF) is the primary method for aligning large language models with human preferences, but it often incurs significant computational and financial costs due to its reliance on training external reward models or human-labeled preferences. |
| Approach: | They propose an alternative approach that leverages generative LLMs as preference classifiers to reduce the dependence on external reward models or human-labeled preferences. |
| Outcome: | The proposed approach reduces the dependence on external reward models or human-labeled preferences by using generative LLMs as preference classifiers. |
COMPACT: Building Compliance Paralegals via Clause Graph Reasoning over Contracts (2026.eacl-long)
Copied to clipboard
| Challenge: | Existing legal NLP benchmarks focus on single-clause tasks, such as ContractNLI and CUAD. |
| Approach: | They propose a framework that models cross-clause dependencies through structured clause graphs by extracting deontic-temporal entities from clauses and constructs typed relationship graphs capturing definitional dependencies, exception hierarchies, and temporal sequences. |
| Outcome: | The proposed framework extracts deontic-temporal entities from clauses and constructs typed relationship graphs capturing definitional dependencies, exception hierarchies, and temporal sequences. |
Samayik: A Benchmark and Dataset for English-Sanskrit Translation (2024.lrec-main)
Copied to clipboard
Ayush Maheshwari, Ashim Gupta, Amrith Krishna, Atul Kumar Singh, Ganesh Ramakrishnan, Anil Kumar Gourishetty, Jitin Singla
| Challenge: | Existing Sanskrit corpora focus on poetry and offer limited coverage of contemporary written materials. |
| Approach: | They release a dataset of 53,000 parallel English-Sanskrit sentences . they use spoken content that covers contemporary world affairs and interpretations . |
| Outcome: | a new dataset of 53,000 parallel English-Sanskrit sentences is released . the dataset outperforms existing models trained on older classical-era poetry datasets . |
The Inner Monologue of Language Models: When Reasoning Traces Reveal More Than They Hide (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in large language models have enabled them to tackle complex tasks . a fundamental question is: are these models aware of what they "learn" and "think"? |
| Approach: | They define three core competencies: awareness of learned latent policies, generalization of these policies across domains, alignment between internal reasoning traces and final outputs. |
| Outcome: | The results show that RL-trained models exhibit stronger generalizability to novel tasks than SFT models but weak alignment between reasoning traces and final outputs. |
COGMEN: COntextualized GNN based Multimodal Emotion recognitioN (2022.naacl-main)
Copied to clipboard
| Challenge: | During a conversation, a person’s emotions are influenced by the other speaker’s utterances and their own emotional state over the utterrances. |
| Approach: | They propose a Graph Neural Network based Multi-modal Emotion recognitioN system that leverages local and global information in a conversation. |
| Outcome: | The proposed system gives state-of-the-art results on IEMOCAP and MOSEI datasets and detailed ablation experiments show the importance of modeling information at both levels. |
LexGen: Domain-aware Multilingual Lexicon Generation (2025.acl-long)
Copied to clipboard
Ayush Maheshwari, Atul Kumar Singh, N J Karthika, Krishnakant Bhatt, Preethi Jyothi, Ganesh Ramakrishnan
| Challenge: | Lexicon generation is a key task in specialized domains due to infrequent usage of terms . a new model is proposed to generate dictionary words for 6 Indian languages . |
| Approach: | They propose a model to generate dictionary words for 6 Indian languages in the multi-domain setting. |
| Outcome: | The proposed model generalizes to unseen domains and unsealed languages. |