{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ASAMY472MDZ3K2YLEO73XOMVWW","short_pith_number":"pith:ASAMY472","schema_version":"1.0","canonical_sha256":"0480cc73fa60f3b56b0b23bfbbb995b5a627cbc09dca28b4c1532858135fb896","source":{"kind":"arxiv","id":"2512.12677","version":2},"attestation_state":"computed","paper":{"title":"Fine-Tuning Causal LLMs for Text Classification: Embedding-Based vs. Instruction-Based Approaches","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Amirhossein Yousefiramandi, Ciaran Cooney","submitted_at":"2025-12-14T13:02:06Z","abstract_excerpt":"We explore efficient strategies to fine-tune decoder-only Large Language Models (LLMs) for downstream text classification under resource constraints. Two approaches are investigated: (1) attaching a classification head to a pre-trained causal LLM and fine-tuning on the task using the LLM's final-token embedding as a sequence representation, and (2) instruction-tuning the LLM in a prompt-to-response format for classification. To enable single-GPU fine-tuning of models up to 8B parameters, we combine 4-bit model quantization with Low-Rank Adaptation (LoRA) for parameter-efficient training. Exper"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2512.12677","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-12-14T13:02:06Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"07b4f42c7c7fd67c6a94237c3a92cf0901ce57aeedb5c677ba14ec68852938c6","abstract_canon_sha256":"bcf6e6008a93594f127c6e2271ce0bd22b7f2e3b8cd02cb969dca5b8414ea48d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-25T02:01:10.843004Z","signature_b64":"r+qMCKJKSp0d6RAWUk+uYQpO+41qOMwK4ZYCGoTO/qCjoK6YcZ4vWyJIed/FfN6XI35nF9f6NMglQVG+r8ECBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0480cc73fa60f3b56b0b23bfbbb995b5a627cbc09dca28b4c1532858135fb896","last_reissued_at":"2026-05-25T02:01:10.842431Z","signature_status":"signed_v1","first_computed_at":"2026-05-25T02:01:10.842431Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fine-Tuning Causal LLMs for Text Classification: Embedding-Based vs. Instruction-Based Approaches","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Amirhossein Yousefiramandi, Ciaran Cooney","submitted_at":"2025-12-14T13:02:06Z","abstract_excerpt":"We explore efficient strategies to fine-tune decoder-only Large Language Models (LLMs) for downstream text classification under resource constraints. Two approaches are investigated: (1) attaching a classification head to a pre-trained causal LLM and fine-tuning on the task using the LLM's final-token embedding as a sequence representation, and (2) instruction-tuning the LLM in a prompt-to-response format for classification. To enable single-GPU fine-tuning of models up to 8B parameters, we combine 4-bit model quantization with Low-Rank Adaptation (LoRA) for parameter-efficient training. Exper"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2512.12677","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2512.12677/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":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2512.12677","created_at":"2026-05-25T02:01:10.842533+00:00"},{"alias_kind":"arxiv_version","alias_value":"2512.12677v2","created_at":"2026-05-25T02:01:10.842533+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2512.12677","created_at":"2026-05-25T02:01:10.842533+00:00"},{"alias_kind":"pith_short_12","alias_value":"ASAMY472MDZ3","created_at":"2026-05-25T02:01:10.842533+00:00"},{"alias_kind":"pith_short_16","alias_value":"ASAMY472MDZ3K2YL","created_at":"2026-05-25T02:01:10.842533+00:00"},{"alias_kind":"pith_short_8","alias_value":"ASAMY472","created_at":"2026-05-25T02:01:10.842533+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.30677","citing_title":"Investigating Detection and Obfuscation of Prompt Injection Attacks Against Software Reverse Engineering AI Agents","ref_index":18,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ASAMY472MDZ3K2YLEO73XOMVWW","json":"https://pith.science/pith/ASAMY472MDZ3K2YLEO73XOMVWW.json","graph_json":"https://pith.science/api/pith-number/ASAMY472MDZ3K2YLEO73XOMVWW/graph.json","events_json":"https://pith.science/api/pith-number/ASAMY472MDZ3K2YLEO73XOMVWW/events.json","paper":"https://pith.science/paper/ASAMY472"},"agent_actions":{"view_html":"https://pith.science/pith/ASAMY472MDZ3K2YLEO73XOMVWW","download_json":"https://pith.science/pith/ASAMY472MDZ3K2YLEO73XOMVWW.json","view_paper":"https://pith.science/paper/ASAMY472","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2512.12677&json=true","fetch_graph":"https://pith.science/api/pith-number/ASAMY472MDZ3K2YLEO73XOMVWW/graph.json","fetch_events":"https://pith.science/api/pith-number/ASAMY472MDZ3K2YLEO73XOMVWW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ASAMY472MDZ3K2YLEO73XOMVWW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ASAMY472MDZ3K2YLEO73XOMVWW/action/storage_attestation","attest_author":"https://pith.science/pith/ASAMY472MDZ3K2YLEO73XOMVWW/action/author_attestation","sign_citation":"https://pith.science/pith/ASAMY472MDZ3K2YLEO73XOMVWW/action/citation_signature","submit_replication":"https://pith.science/pith/ASAMY472MDZ3K2YLEO73XOMVWW/action/replication_record"}},"created_at":"2026-05-25T02:01:10.842533+00:00","updated_at":"2026-05-25T02:01:10.842533+00:00"}