eng-ml
It would enforce do_sample=False, temperature=0, top_k=1, and generator usage. Name it deterministic=True.
dsid_1d487c52d79a44c2b40eb19a0db51122
qst_0415
deterministic true enforcement
do_sample=false, temperature=0, top_k=1, explicit generator
also top_p=1.0, num_beams=1, stable batch order
Exact spans from the benchmark corpus
eng-ml
It would enforce do_sample=False, temperature=0, top_k=1, and generator usage. Name it deterministic=True.
dsid_1d487c52d79a44c2b40eb19a0db51122
CI Eval Determinism and Flake Mitigation
When deterministic: true, the harness MUST enforce do_sample=false, temperature=0.0, top_k=1, top_p=1.0, num_beams=1, explicit generator, stable batch order.
dsid_05841423a55e4c35abd276d9899b7134
Every document that still carries the retired value, including the superseded source itself
No residue recorded for this claim. Zero findings are shown as zero.
Every query that produced this page
MATCH (c:ClaimKey {question_id: $question_id}) RETURN c.id AS id, c.entity_id AS entity_id, c.key AS key, c.predicate AS predicate, c.question_id AS question_idMATCH (e:Entity)-[:HAS_CLAIM]->(c:ClaimKey {id: $claim_key_id}) RETURN e.id AS id, e.name AS name, e.entity_type AS entity_typeMATCH (c:ClaimKey {id: $claim_key_id})-[:HAS_VALUE]->(p:Proposition) RETURN p.id AS id, p.value AS value, p.status AS status ORDER BY p.idMATCH (c:ClaimKey {id: $claim_key_id})-[:HAS_VALUE]->(p:Proposition)<-[:SELECTS]-(ev:CanonEvent) OPTIONAL MATCH (newer:CanonEvent)-[:SUPERSEDES]->(ev) RETURN ev.id AS id, ev.claim_key_id AS claim_key_id, ev.selects_proposition_id AS selects_proposition_id, ev.transition AS transition, ev.temporal_quality AS temporal_quality, ev.evidence_doc_id AS evidence_doc_id, ev.occurred_at AS occurred_at, p.id AS proposition_id, newer.id AS superseded_byMATCH (ev:CanonEvent {id: $event_id})-[:SUPERSEDES*1..10]->(old:CanonEvent)-[:SELECTS]->(p:Proposition) RETURN old.id AS id, old.transition AS transition, old.temporal_quality AS temporal_quality, old.evidence_doc_id AS evidence_doc_id, old.occurred_at AS occurred_at, p.id AS proposition_id, p.value AS valueMATCH (p:Proposition {id: $proposition_id})<-[:ASSERTS]-(a:Assertion)-[:IN_ARTIFACT]->(d:Artifact) RETURN a.id AS id, a.doc_id AS doc_id, a.source_type AS source_type, a.evidence_span AS evidence_span, a.stance AS stance, a.extraction_method AS extraction_method, a.structured AS structured, a.discovery AS discovery, a.asserted_at AS asserted_at, a.source_field AS source_field, a.residue_class AS residue_class, a.proposition_id AS proposition_id, a.artifact_id AS artifact_id, d.title AS title ORDER BY a.doc_idMATCH (p:Proposition {id: $proposition_id})<-[:ASSERTS]-(a:Assertion)-[:IN_ARTIFACT]->(d:Artifact) RETURN a.id AS id, a.doc_id AS doc_id, a.source_type AS source_type, a.evidence_span AS evidence_span, a.stance AS stance, a.extraction_method AS extraction_method, a.structured AS structured, a.discovery AS discovery, a.asserted_at AS asserted_at, a.source_field AS source_field, a.residue_class AS residue_class, a.proposition_id AS proposition_id, a.artifact_id AS artifact_id, d.title AS title ORDER BY a.doc_idMATCH (p:Proposition {id: $proposition_id})<-[:ASSERTS]-(a:Assertion)-[:IN_ARTIFACT]->(d:Artifact) RETURN a.id AS id, a.doc_id AS doc_id, a.source_type AS source_type, a.evidence_span AS evidence_span, a.stance AS stance, a.extraction_method AS extraction_method, a.structured AS structured, a.discovery AS discovery, a.asserted_at AS asserted_at, a.source_field AS source_field, a.residue_class AS residue_class, a.proposition_id AS proposition_id, a.artifact_id AS artifact_id, d.title AS title ORDER BY a.doc_id