{"evaluation_date":"2026-09-26","kev_model":"jaredpalmer/kev-4b","kev_revision":"139fdd94f1b6a6ad80cc15e08fcb99cac885a101","kev_code_revision":"58d94380d4441d2d9fbb7e5e6d30d9a5b93578ae","clm_harness_revision":"7956937c58ed5839c06ddc4dc6b6b61c3a3e4094","protocol":"Original fixed 40 tickets twice; 369 narrow decisions; 1253 BFCL cases in description, schema and prose_schema views; T-Rex seeds 0-4, 60s, 6 in flight, original course, labeled prompt, shield on/off. Fresh Jev calls; no fine-tuning or date preprocessing. Kev official bf16 fused serving defaults and checkpoint temperature.","compute_host":{"id":"jutgzx3z3xhonn","gpu":"NVIDIA RTX PRO 4500 Blackwell Server Edition","datacenter":"US-KS-2","gpu_usd_per_hour":0.72,"allocated_at":"2026-09-26T07:56:56Z","container_disk_gb":50,"volume_gb":50},"fixture_sha256":{"evaluation/agent_claims/fixtures.py":"82fe25f1ecda0420ddd8e72a433114bd4b374f83785d6e0ea0227c015ca95d34","evaluation/bfcl_v4/source/BFCL_v4_live_multiple_answers.json":"97e90d59c5bd76c55a2920ce93e5566e9046307d3f558578f085f9d3a56c3084","evaluation/bfcl_v4/source/BFCL_v4_multiple_answers.json":"244e00ce9395df948bcafc7bee64e8f9c87ef70887587d83cae45b13699f3047","evaluation/bfcl_v4/source/BFCL_v4_live_multiple.json":"fd8ccfad4d911420d0e3341dbe2fff77d1d341da934248b9bb2bda24ab3a10c8","evaluation/bfcl_v4/source/BFCL_v4_multiple.json":"aef168155ebd74b7ac2401198b201343bc7d16d7a3d7e0d4e6d8ee82c6969b2a"},"report_deployment":{"platform":"Cloudflare Workers Static Assets","domain":"kev-eval.oaleinyk.xyz","worker_script":false},"issues":[{"stage":"bfcl_prose_schema_jev","error":"Hosted Jev API returned HTTP 520 after at least 400 successful cases; the old retry allowlist did not include 520.","resolution":"Added HTTP 520 to bounded retry handling, resumed the ledger, reran only the interrupted full view, and preserved completed runs. The interrupted attempt is excluded from reported scores and token totals."}],"probability_precision_note":{"suite":"bfcl_prose_schema_jev","case":"live_multiple_44-17-0","sum":0.99,"precision":2,"handling":"Original probabilities preserved without renormalization; validation permits the mathematically bounded rounding error."},"source_result_sha256":{"complete.json":"9a6a891ca46a04fdaa45c2bdab244123247fe5c0aca2c70040b9c552f3f47a96","server-final.json":"1b82e86ceadf3c6958209787ddbeeb420c16b63bd03fbccbab5af94474a0a900","policy_jev_shield.json":"3780e5f753248fe976ad17edde91cbf0dc17a51f1b9b9086488ca8ece7baa969","policy_kev_shield.json":"43713cb0aa2568c1a079a87a641abba3644d6fd49983253484cf16d5f4aed2a9","policy_jev_noshield.json":"cdec61e499f070875a2b93123297afcfd683cb8b6bbf3d3e9407cc583f20f17f","policy_kev_noshield.json":"d85495d6987590cdadb717f9cc0a0167a7658a798337d952d645be536d0b9f8b","bfcl_prose_schema_jev.json":"f44877ef7483db0f6f8b969c98b16399c4a8b68fbe9bedf5c5de6d4aa2f4313d","bfcl_prose_schema_kev.json":"09b87644304087412fc68c76ffcbc09eb20b72e91eab646ebb80a215c7757679","server-prose_schema.json":"c012ef0685250964e7ac4daf21b681a2df63f0eefb7dfb1ef2e8b2ac148b78d9","bfcl_schema_jev.json":"f73ac13c010f41227768b3f4b8cb26057d7f77040f3ae1a338f331da6a022f77","bfcl_schema_kev.json":"7c3c1a8b60878602bda64e330d1df355a5c11aaabe38df59f20c54d7c6ecc9a0","server-schema.json":"c012ef0685250964e7ac4daf21b681a2df63f0eefb7dfb1ef2e8b2ac148b78d9","bfcl_description_jev.json":"e83bc4bfd0adaa345b0dd2f94cb8f33cad963a4170b824beefc4d8700c72c584","bfcl_description_kev.json":"08098876d62e55dcab074800d6b0ab1eb7645e6778827cf052d62f8799f6e6e6","server-description.json":"c012ef0685250964e7ac4daf21b681a2df63f0eefb7dfb1ef2e8b2ac148b78d9","narrow_jev.json":"624172a94f739263605c96de9a05e05c9780966331a4c9d67ee861fd7c07eef9","tickets_jev.json":"2a5a4761158178cdc3be582df3375d0a6b6cced34dc08a1197bd3f0df36ff0f9","narrow_kev.json":"307a9b8882c8b0d9a334f86adde7d8691f2f8419eb7eccc0ce4f853fd13dd514","tickets_kev.json":"7f421405fe05a6fc80e7a8da8b0f418c6a994b88e7cfd28d277066bb7a16635d","run-ledger.json":"4b43c53648a533c92b137a8ea15280b73ea785a17a512a6ba957425877caad60","server-initial.json":"c012ef0685250964e7ac4daf21b681a2df63f0eefb7dfb1ef2e8b2ac148b78d9","environment.json":"e9661cbd3c1b23c618ef0282d71bd9551767afd38f045f345ac0a4c1cde7bdb8"},"cleanup":{"verified_empty":true,"deleted_pod":"jutgzx3z3xhonn","deleted_at":"2026-09-26T08:54:12Z","verified_at":"2026-09-26T08:54:32.402749+00:00","pods":[],"serverless_endpoints":[],"network_volumes":[],"estimated_gpu_cost_usd":0.6872,"cost_note":"Elapsed allocation at quoted GPU rate; excludes disk billing. Not an invoice."},"publication_note":"Compute metadata uses generic labels. Source checksums identify the original saved evaluation artifacts."}
