{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QP5TDKJVMYRBQPCZLU6I4DA7PQ","short_pith_number":"pith:QP5TDKJV","schema_version":"1.0","canonical_sha256":"83fb31a9356622183c595d3c8e0c1f7c12e728e8e425dd99edde67971f472dfa","source":{"kind":"arxiv","id":"2606.25782","version":1},"attestation_state":"computed","paper":{"title":"Do Encoders Suffice? A Systematic Comparison of Encoder and Decoder Safety Judges for LLM Adversarial Evaluation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Han Jeon, Joseph Voyles, Matt Wood, Shiv Medler","submitted_at":"2026-06-24T13:00:25Z","abstract_excerpt":"With the widespread adoption of large language models (LLMs) in chatbots and everyday applications, companies increasingly need guardrails that are effective while remaining low-cost and low-latency. Safety evaluation of LLM outputs has generally relied on LLM-based judges, which can be effective but are often slow and expensive to deploy at scale.\n  In this paper, we evaluate whether fine-tuned modern encoder classifiers from the ModernBERT family, including ModernBERT and Ettin, can reliably identify harmful LLM outputs in user-model conversations without substantial performance loss relativ"},"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":"2606.25782","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-24T13:00:25Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a5ab0f2b6540c3ff7de63b537b8208c386e91da5fd00bbf90a8730d66d664ebd","abstract_canon_sha256":"327b843ee537e47e886fb20e63c01ac164bdac703a6f493bcf575f78fb044b96"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-25T01:18:15.018860Z","signature_b64":"rwSfCKEQCMdE8vDEtzxqMaQtyLX482JcxaNO7EhaBKqFBycSU5BSsy1ObiXN37RKJzMD59jNJV0z298CW8qUBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"83fb31a9356622183c595d3c8e0c1f7c12e728e8e425dd99edde67971f472dfa","last_reissued_at":"2026-06-25T01:18:15.018515Z","signature_status":"signed_v1","first_computed_at":"2026-06-25T01:18:15.018515Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Do Encoders Suffice? A Systematic Comparison of Encoder and Decoder Safety Judges for LLM Adversarial Evaluation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Han Jeon, Joseph Voyles, Matt Wood, Shiv Medler","submitted_at":"2026-06-24T13:00:25Z","abstract_excerpt":"With the widespread adoption of large language models (LLMs) in chatbots and everyday applications, companies increasingly need guardrails that are effective while remaining low-cost and low-latency. Safety evaluation of LLM outputs has generally relied on LLM-based judges, which can be effective but are often slow and expensive to deploy at scale.\n  In this paper, we evaluate whether fine-tuned modern encoder classifiers from the ModernBERT family, including ModernBERT and Ettin, can reliably identify harmful LLM outputs in user-model conversations without substantial performance loss relativ"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.25782","kind":"arxiv","version":1},"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/2606.25782/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":"2606.25782","created_at":"2026-06-25T01:18:15.018579+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.25782v1","created_at":"2026-06-25T01:18:15.018579+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.25782","created_at":"2026-06-25T01:18:15.018579+00:00"},{"alias_kind":"pith_short_12","alias_value":"QP5TDKJVMYRB","created_at":"2026-06-25T01:18:15.018579+00:00"},{"alias_kind":"pith_short_16","alias_value":"QP5TDKJVMYRBQPCZ","created_at":"2026-06-25T01:18:15.018579+00:00"},{"alias_kind":"pith_short_8","alias_value":"QP5TDKJV","created_at":"2026-06-25T01:18:15.018579+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QP5TDKJVMYRBQPCZLU6I4DA7PQ","json":"https://pith.science/pith/QP5TDKJVMYRBQPCZLU6I4DA7PQ.json","graph_json":"https://pith.science/api/pith-number/QP5TDKJVMYRBQPCZLU6I4DA7PQ/graph.json","events_json":"https://pith.science/api/pith-number/QP5TDKJVMYRBQPCZLU6I4DA7PQ/events.json","paper":"https://pith.science/paper/QP5TDKJV"},"agent_actions":{"view_html":"https://pith.science/pith/QP5TDKJVMYRBQPCZLU6I4DA7PQ","download_json":"https://pith.science/pith/QP5TDKJVMYRBQPCZLU6I4DA7PQ.json","view_paper":"https://pith.science/paper/QP5TDKJV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.25782&json=true","fetch_graph":"https://pith.science/api/pith-number/QP5TDKJVMYRBQPCZLU6I4DA7PQ/graph.json","fetch_events":"https://pith.science/api/pith-number/QP5TDKJVMYRBQPCZLU6I4DA7PQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QP5TDKJVMYRBQPCZLU6I4DA7PQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QP5TDKJVMYRBQPCZLU6I4DA7PQ/action/storage_attestation","attest_author":"https://pith.science/pith/QP5TDKJVMYRBQPCZLU6I4DA7PQ/action/author_attestation","sign_citation":"https://pith.science/pith/QP5TDKJVMYRBQPCZLU6I4DA7PQ/action/citation_signature","submit_replication":"https://pith.science/pith/QP5TDKJVMYRBQPCZLU6I4DA7PQ/action/replication_record"}},"created_at":"2026-06-25T01:18:15.018579+00:00","updated_at":"2026-06-25T01:18:15.018579+00:00"}