FA-98626 / KV cache position mapping / Open access
Multi-token append reserves capacity atomically · case 01
A request that does not fit partially consumes free cache slots.
ROOT CAUSE
Each token is admitted independently before the full sequence capacity is known.
THE FAILURE
Each token is admitted independently before the full sequence capacity is known.
Unsuccessful approach: Comparing the request to total capacity ignores slots already occupied and can admit too many tokens.
Case contract
Return the requested count if it fits in capacity-used, else zero; never return a partial count.
Why this case matters
A small in-memory model makes the cache address contract explicit for a serving component.
1 / The failure
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(capacity, used, requested):
return min(requested,capacity-used)
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('fits exactly',solve(N+3,N,N+3-N),N+3-N)
check('one free short',solve(N+3,N+1,3),0)
check('partial would fit',solve(5,3,4),0)
check('empty append',solve(5,3,0),0)
check('empty cache',solve(N,0,N),N)
check('full cache',solve(N,N,1),0)
check('negative request',solve(5,0,-1),0)
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 fixture | Actual | Expected | Outcome |
|---|---|---|---|
| fits exactly | 3 | 3 | Passed |
| one free short | 2 | 0 | Failed |
| partial would fit | 2 | 0 | Failed |
| empty append | 0 | 0 | Passed |
| empty cache | 1 | 1 | Passed |
| full cache | 0 | 0 | Passed |
| negative request | -1 | 0 | Failed |
SHA-256 / f0d709fa59ddb71be57562a4d7a7f871b1839700a6ccbef359a072b083c69be7
2 / The unsuccessful fix
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(capacity, used, requested):
return requested if requested<=capacity else 0
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('fits exactly',solve(N+3,N,N+3-N),N+3-N)
check('one free short',solve(N+3,N+1,3),0)
check('partial would fit',solve(5,3,4),0)
check('empty append',solve(5,3,0),0)
check('empty cache',solve(N,0,N),N)
check('full cache',solve(N,N,1),0)
check('negative request',solve(5,0,-1),0)
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 fixture | Actual | Expected | Outcome |
|---|---|---|---|
| fits exactly | 3 | 3 | Passed |
| one free short | 3 | 0 | Failed |
| partial would fit | 4 | 0 | Failed |
| empty append | 0 | 0 | Passed |
| empty cache | 1 | 1 | Passed |
| full cache | 1 | 0 | Failed |
| negative request | -1 | 0 | Failed |
SHA-256 / a5c2b9696eb36f7c6302d0740d7a2dc55b0bf6e77e19301345381e84275301d9
HELD IN THE MEMBER ARCHIVE
The verified repair and its recorded checks are member-only.
This mechanism has 7 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
These toy integer maps omit tensor layout, concurrency, device memory, and model-specific cache formats. 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 2026-09-30T16:48:14.972284+00:00.
Case digest / 4e01a54efd3e9e4a8d2d8d6392e2cb30cb3826fa2bf71bd0f4afebdc3e1ff4df