{"paper":{"title":"From Fragments to Facts: A Curriculum-Driven DPO Approach for Generating Hindi News Veracity Explanations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A curriculum-driven DPO framework generates reliable Hindi news veracity explanations by preferring fact-checked sources.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Adam Jatowt, Pulkit Bansal, Raghvendra Kumar, Shakti Singh, Sriparna Saha","submitted_at":"2025-07-07T16:34:28Z","abstract_excerpt":"In an era of rampant misinformation, generating reliable news explanations is vital, especially for under-represented languages like Hindi. Lacking robust automated tools, Hindi faces challenges in scaling misinformation detection. To bridge this gap, we propose a novel framework integrating Direct Preference Optimization (DPO) with curriculum learning to align machine-generated explanations with human reasoning. Fact-checked explanations from credible sources serve as preferred responses, while LLM outputs highlight system limitations and serve as non-preferred responses. To refine task-speci"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments with LLMs (Mistral, Llama, Gemma) and PLMs (mBART, mT5) confirm the framework's effectiveness in generating coherent, contextually relevant explanations.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That fact-checked explanations from credible sources can reliably serve as preferred responses while LLM outputs serve as non-preferred responses, and that adding the Actuality and Finesse parameters to the DPO loss will produce measurable improvements in explanation quality without introducing new biases or inconsistencies.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A DPO framework augmented with curriculum learning and two new loss parameters generates veracity explanations for Hindi news using LLMs and PLMs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A curriculum-driven DPO framework generates reliable Hindi news veracity explanations by preferring fact-checked sources.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"9223b97f260b554eca0cbb36ee18200db00605ddb6e852a8b26544a2b50d6c62"},"source":{"id":"2507.05179","kind":"arxiv","version":6},"verdict":{"id":"2ce27f78-c6e8-44ec-8fc4-003038fbf60c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T06:04:03.724054Z","strongest_claim":"Experiments with LLMs (Mistral, Llama, Gemma) and PLMs (mBART, mT5) confirm the framework's effectiveness in generating coherent, contextually relevant explanations.","one_line_summary":"A DPO framework augmented with curriculum learning and two new loss parameters generates veracity explanations for Hindi news using LLMs and PLMs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That fact-checked explanations from credible sources can reliably serve as preferred responses while LLM outputs serve as non-preferred responses, and that adding the Actuality and Finesse parameters to the DPO loss will produce measurable improvements in explanation quality without introducing new biases or inconsistencies.","pith_extraction_headline":"A curriculum-driven DPO framework generates reliable Hindi news veracity explanations by preferring fact-checked sources."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2507.05179/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":1,"snapshot_sha256":"48a212deabcf3159a1b8e2f3a7419ec5b21e20f66ccd7d1731cbee01a61a6932"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}