correct stale cost paragraph and smoke docstring to committed design

This commit is contained in:
wassname committed 2026-09-19 17:06:16 +08:00
1 parent b3b394b227
commit 703e37156b
2 files changed
+2 -2

No files matched your search

+1 -1
View File
@@ -1208,7 +1208,7 @@ One API call is ONE pseudo-respondent answering the complete selected battery in
### Design, request count, cost
N=128 whole respondent packets per release with 128 paired deterministic seeds shared across releases (same seed per packet index on every release); canonical option order (no rotation; the packet is one coherent questionnaire). Request count: 4 releases x 128 packets = 512 panel calls plus 1 paid smoke = 513. Reserve bound per packet: 1,024 input + 2,048 output tokens at the saved Alibaba prices (prompt 0.26-0.325 and completion 1.28-1.95 USD per million): panel bound USD 3.770941440, smoke USD 0.00692224, total USD 3.777863680 against the USD 5 stage stop, all inside the existing locked USD 80 repository ledger under lane `google` with reservation `pilot/resp-packet`.
N=128 whole respondent packets per release with 128 paired deterministic seeds shared across releases (same seed per packet index on every release); canonical option order (no rotation; the packet is one coherent questionnaire). Request count: 4 releases x 128 packets = 512 panel calls plus 1 paid smoke = 513. Reserve bound per API phase: 1,024 input + 2,048 output tokens at the saved Alibaba prices: panel initial-call bound USD 1.885470720, smoke bound USD 0.00294912 on qwen/qwen3.7-plus, total USD 1.888419840; each rescue/retry phase reserves another per-phase bound, and the USD 5 stage stop (checked on every reserve) is the authoritative all-in limit, all inside the existing locked USD 80 repository ledger under lane `alibaba` with reservation `pilot/resp-packet-smoke` for the smoke. (An earlier draft of this paragraph quoted doubled bounds and the wrong lane; the manifest and code values above govern.)
### Analysis (fixed before unblinding)
+1 -1
View File
@@ -847,7 +847,7 @@ def offline_regression(battery: list[dict]) -> None:
def paid_smoke(battery: list[dict]) -> None:
"""One paid packet on qwen/qwen3.7-plus (default reasoning enabled: riskiest parse path)."""
"""One paid packet on qwen/qwen3.7-plus with reasoning disabled, matching the panel payload."""
model = "qwen/qwen3.7-plus"
pid = protocol_id(model, battery)
req = {"packet": 0, "seed": paired_seeds(N_PACKETS)[0], "prompt": render_packet(battery)}