This paper discusses the Retrieval-Augmented Generation (RAG) system submitted to the MTRAG-UN shared task on multi-turn conversational question answering. The paper describes the proposed solution for Task A (Document Retrieval) and Task C (Full RAG Pipeline), focusing on retrieval robustness and grounded response generation in complex English multi-turn dialogs. The proposed retrieval architecture uses a cascaded hybrid pipeline, which combines sparse retrieval (BM25) with dense bi-encoder models (BGE-base-en-v1.5 and E5-base), integrated via Reciprocal Rank Fusion and refined using a weighted ensemble of cross-encoders. For the generation part, the top-3 retrieved passages are injected into FLAN-T5-Large using an anchor-prompting strategy to output grounded faithful responses. Experimental results show that the proposed hybrid retrieval framework with multi-stage reranking significantly enhances passage selection, particularly for non-standalone conversational queries. Further analysis reveals persistent difficulties in handling underspecified and unanswerable questions, as well as an increased susceptibility to retrieval noise in later dialog turns.

Words are powerful; they shape thoughts that influence actions and reveal emotions. On social media, where billions of people share theiropinions daily. Comments are the key to understanding how users feel about a video, an image, or even an idea. But what happens when these comments are messy, riddled with code-mixed language, emojis, and informal text? The challenge becomes even greater when analyzing low-resource languages like Tamil and Tulu. To tackle this, TensorTalk deployed cutting-edge machine learning techniques such as Logistic regression for Tamil language and SVM for Tulu language , to breathe life into unstructured data. By balancing, cleaning, and processing comments, TensorTalk broke through barriers like transliteration and tokenization, unlocking the emotions buried in the language.