FA-98485 / Beam search bookkeeping / Member archive
Beam search: length penalty counts prompt tokens as generated output · case 05
length penalty counts prompt tokens as generated output.
Case contract
Rows [id,score,total_length]; normalize by max(1,total_length-prompt_length) and return scores by ID.
Why this case matters
Beam decoders rely on consistent hypothesis state while expanding and ranking candidates.
One recorded failure
Sample boundary fixtureThis sample comes from the broken implementation of a controlled reproducer.
| Boundary fixture | Actual | Expected | Outcome |
|---|---|---|---|
| shared prompt removed | {"a": 0.7142857142857143} | {"a": 2.5} | Failed |
MEMBER ARCHIVE
The complete case is available to members.
This record includes three runnable implementations, regression fixtures, execution results, and source hashes.
Member access is invitation-based. Sign in with your invited account to inspect the sources.
Sign in to the archive ↗