Papers by Souvik Das
UNIWIZ: A Unified Large Language Model Orchestrated Wizard for Safe Knowledge Grounded Conversations (2024.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have made significant progress in integrating safety and knowledge alignment, but excessive focus on safety alignment can lead to unintended hallucinations. |
| Approach: | They propose a "safety-priming" method to generate synthetic safety data and overcome safety bottlenecks. |
| Outcome: | The proposed framework generates synthetic safety data and overcomes safety bottlenecks. |
Entropy Guided Extrapolative Decoding to Improve Factuality in Large Language Models (2025.coling-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) exhibit impressive natural language capabilities but suffer from hallucination – generating content that does not align with realworld facts. |
| Approach: | They propose to extrapolate critical token probabilities beyond the last layer to improve decoding by manipulating the predicted distributions at inference time. |
| Outcome: | The proposed methods surpass state-of-the-art on multiple datasets by large margins. |
Using Multi-Encoder Fusion Strategies to Improve Personalized Response Selection (2022.coling-1)
Copied to clipboard
| Challenge: | Existing systems that focus on persona do not explore well the correlation between persona and empathy. |
| Approach: | They propose a suite of fusion strategies that capture interaction between persona, emotion, and entailment information of the utterances. |
| Outcome: | The proposed model outperforms the previous methods by 2.3% on original personas and 1.9% on revised persona models in terms of hits@1 accuracy. |
Dialo-AP: A Dependency Parsing Based Argument Parser for Dialogues (2022.coling-1)
Copied to clipboard
| Challenge: | a recent work on argument mining has focused on parsing monologues, while neglecting dialogues. |
| Approach: | They propose an end-to-end argument parser that constructs argument graphs from dialogues . they use extensive pre-training and curriculum learning to train AM . |
| Outcome: | The proposed system performs all sub-tasks of AM and achieves significant improvements . it is compared to existing systems and validated through human evaluation . |
Beyond Discrete Personas: Personality Modeling Through Journal Intensive Conversations (2025.coling-main)
Copied to clipboard
| Challenge: | Existing LLMs rely on static, predefined personas to capture dynamic and evolving nature of human personalities. |
| Approach: | They propose a dataset with 400,000 conversations and a framework for generating personalized conversations using long-form journal entries from Reddit. |
| Outcome: | The proposed framework generates high-quality, personality-rich dialogues grounded in reddit journal entries. |
Diving Deep into Modes of Fact Hallucinations in Dialogue Systems (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Knowledge Graph(KG) grounded conversations often use large pre-trained models and suffer from fact hallucination. |
| Approach: | They propose to use a human feedback analysis to identify various modes of hallucination in KG chatbots. |
| Outcome: | The proposed system provides fine-grained signals that control fallacious content while generating responses. |