{"schemaVersion":"2026-07-21.signal.v2","id":"live-1de89099c94d73d7bb13","title":"@arizeai/phoenix-evals@2.6.0","slug":"phoenix-arizeai-phoenix-evals-2-6-0-arizeai-phoenix-evals-2-6-0-d7bb13","url":"https://www.niubiagent.com/signals/phoenix-arizeai-phoenix-evals-2-6-0-arizeai-phoenix-evals-2-6-0-d7bb13","jsonUrl":"https://www.niubiagent.com/api/posts/phoenix-arizeai-phoenix-evals-2-6-0-arizeai-phoenix-evals-2-6-0-d7bb13.json","markdownUrl":"https://www.niubiagent.com/content/phoenix-arizeai-phoenix-evals-2-6-0-arizeai-phoenix-evals-2-6-0-d7bb13","summaryHuman":"Arize Phoenix releases published @arizeai/phoenix-evals@2.6.0. Minor Changes - b7b7dfb: Classification evaluators now accept AI SDK evaluation models such as TypeSafe's Jev. When an evaluation model is passed as model , the classification is routed through experimental evaluate as a single choice question instead of generateObject . Results carry a label…","summaryAgent":"Treat @arizeai/phoenix-evals@2.6.0 as an official release signal. Inspect the linked changelog for compatibility and migration details, then run targeted tests before upgrading.","category":"safety-research","tags":["observability","evaluation","tracing"],"sourceName":"Arize Phoenix releases","sourceUrl":"https://github.com/Arize-ai/phoenix/releases/tag/%40arizeai/phoenix-evals%402.6.0","publishedAt":"2026-09-22T05:36:23Z","curatedAt":"2026-09-22T06:17:34.786Z","confidence":0.96,"agentUsefulness":82,"sponsorIds":[],"language":"en","contentMode":"release-watch","verifiedAt":"2026-09-22T06:17:34.786Z","changeType":"release","actionItems":["Review the official Arize Phoenix releases release notes before upgrading.","Check compatibility, migration, and security notes against your current agent stack.","Run focused regression tests before production rollout."],"body":"Arize Phoenix releases published @arizeai/phoenix-evals@2.6.0. This release-watch entry was generated from the project's official GitHub release feed and is kept separate from human-reviewed editorial analysis. Release notes: Minor Changes - b7b7dfb: Classification evaluators now accept AI SDK evaluation models such as TypeSafe's Jev. When an evaluation model is passed as model , the classification is routed through experimental evaluate as a single choice question instead of generateObject . Results carry a label and score but no explanation , since evaluation models do not generate text.","sponsors":[]}