FAILURE MAP
← Case archive

FA-100781 / Model output stopping rules / Open access

Output stopping: EOS is classified without truncating output · case 01

EOS is classified without truncating output.

Verified by executionVariant 1 · 6 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

The end token is classified but output is not cut at its first occurrence.

THE FAILURE

The end token is classified but output is not cut at its first occurrence.

Unsuccessful approach: Including EOS in the returned prefix exposes the marker as output.

Case contract

Truncate before the first end token and report reason eos; if absent preserve all and report null.

Why this case matters

A deterministic output-control model isolates one finish or streaming boundary from any particular runtime.

1 / The failure

Exit 1
"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(tokens, eos_id):
    return (tokens,None) if eos_id not in tokens else (tokens,'eos')
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('eos middle',solve([1,9,2],9),([1],'eos'))
check('eos first',solve([9],9),([],'eos'))
check('absent',solve([1,2],9),([1,2],None))
check('empty',solve([],9),([],None))
check('repeated eos',solve([1,9,9],9),([1],'eos'))
check('zero eos valid',solve([2,0,3],0),([2],'eos'))
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)
Boundary fixtureActualExpectedOutcome
eos middle[[1, 9, 2], 'eos'][[1], 'eos']Failed
eos first[[9], 'eos'][[], 'eos']Failed
absent[[1, 2], None][[1, 2], None]Passed
empty[[], None][[], None]Passed
repeated eos[[1, 9, 9], 'eos'][[1], 'eos']Failed
zero eos valid[[2, 0, 3], 'eos'][[2], 'eos']Failed

SHA-256 / a3151b4206105282d82fc8093ad48f41d80a310b7a58827ac51fb459fdfd4ab9

2 / The unsuccessful fix

Exit 1
"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(tokens, eos_id):
    if eos_id in tokens:return tokens[:tokens.index(eos_id)+1],'eos'
    return tokens,None
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('eos middle',solve([1,9,2],9),([1],'eos'))
check('eos first',solve([9],9),([],'eos'))
check('absent',solve([1,2],9),([1,2],None))
check('empty',solve([],9),([],None))
check('repeated eos',solve([1,9,9],9),([1],'eos'))
check('zero eos valid',solve([2,0,3],0),([2],'eos'))
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)
Boundary fixtureActualExpectedOutcome
eos middle[[1, 9], 'eos'][[1], 'eos']Failed
eos first[[9], 'eos'][[], 'eos']Failed
absent[[1, 2], None][[1, 2], None]Passed
empty[[], None][[], None]Passed
repeated eos[[1, 9], 'eos'][[1], 'eos']Failed
zero eos valid[[2, 0], 'eos'][[2], 'eos']Failed

SHA-256 / 4cb7c824d6b4a8bed61881e64dea51faf98ef0ef89bc2371fb1efc0a8e3bff81

HELD IN THE MEMBER ARCHIVE

The verified repair and its recorded checks are member-only.

This mechanism has 6 recorded checks per implementation. The open-access tier publishes the failure and the unsuccessful fix; the repaired source that passes every check, and the observations that prove it, are available to members.

Every case sharing this mechanism uses the same contract and the same repair, so this one record is held back for all of them.

Member access is invitation-based. Sign in with your invited account to inspect the repair.

Sign in to the archive ↗

Verification & scope

This toy contract covers the supplied token records only; it does not claim compatibility with a particular model runtime or tokenizer. This reproducer isolates one failure mechanism. Results cover the supplied fixtures. Variants within a family share a test contract and should remain grouped when constructing evaluation splits. Related mechanisms with a shared evaluation_group must also remain together; these controlled models are not independent production incidents.

Observations recorded using Python 3.12.14 at 2026-09-30T16:48:39.593556+00:00.

Case digest / fae43f8ebe4dcd438245f50a2761f063ab246d3dc87bc9de30c876b1111a42a9