Are Transformers Effective for Time Series Forecasting?
The attempt was finalized (its sandbox was taken down) without a finish event and before any result was recorded. It reports no result and makes no finding about the paper.
Plan. Registered before results were recorded; not sealed (legacy).
- Scored claims
- 0 of 0 consistent
- Deviations
- 0 (0 material)
- Integrity flags
- 0
- Logged compute
- 0.01 h over 12 commands and 0 jobs
- Work tree when the attempt ended
- cure-lab/LTSF-Linear@721ab5879ba7 in /work/src-721ab58
- Work tree when the attempt ended
- cure-lab/LTSF-Linear@6f6cce8bc8fa in /work/src-6f6cce8
Claims
Deviations from the paper's protocol
None recorded.
Notes from the attempt
- event 22
Setup-only attempt, abandoned before any claim registration, seal, training or inference. The operator ran environment/data/source setup (events 5-21) before registering the claims, which the engine counts as claims registered after observed output. The settling test is restarted in a fresh attempt that registers all eight claims before any command output; this attempt holds no results and is not part of the ledger.
Assurance
What the harness enforced for this attempt, and what it predates. A legacy attempt is labelled, never rewritten.
- Plan sealed
- no
- Sealed before scored execution
- no
- Evidence bound to artifact excerpts
- no
- Values extracted by a sealed rule
- no
- Execution manifest for cited jobs
- no
- Semantic re-derivation passed
- yes
- Engine
- report built by repro 0.2.1; attempt created by repro 0.2.1
Redaction before release
Before release, the operator's local home-directory path was replaced with ~ in 4 places. Every event hash from event 16 on changed with it. The redaction records the outcome, verdicts and values from before it. That record is an operator assertion, checked for internal consistency against the current log, not proof that the original log was unchanged: the original is not part of the record. It matches what the log derives now.
Rule: operator home directory prefix → ~. Reason: Before release: the operator's local home-directory path, copied from a settling-plan file, was replaced with ~ Recorded at event 26; the head the operator states for the log before the redaction is cce25ce9e195c25923377e98d67e628fb2cec16670b76bea94a064a74c3140c9 (not checkable from this log; a copy anchored elsewhere or the operator's kept original can check it).
Checked from the record: the redacted log's hash chain and artifact digests; re-derivation of every judgement from it; the redaction event's internal consistency (its position, replacement rows, recomputed digests and plan hashes, a single redaction or a second that names the first); and that the outcome, qualifiers, tallies, assurance flags, verdicts and observed values it records as before equal what the log derives now. Not checked: any other text (metric and protocol wording, units, paths), what files and commands meant when they ran, or anything about the log before the redaction, which is not in the record.
Provenance
- Driven by
- operator:claude-opus-5-5 (Claude Code subagent, AI-driven)
- Sandbox
- 3.5 CPUs, 6.00 GiB memory, no GPU · image repro-sandbox:0.2
- Period
- 2026-10-20T02:48:00.615Z → 2026-10-20T02:56:12.284Z
- Event log
- 27 events, hash-chained; head 65c36f2558bfab544aa3eb44e5863236d5784ccb27fa908d5313ac720c87b2d8
- Verify
- python3 openrepro_verify.py arxiv-2205-13504-20261020T024800Z checks the chain and every artifact with the Python standard library alone; repro verify also recomputes every quote match, evidence check and verdict and re-derives this report (engine 0.2.1) (the engine that does this is not yet public; the standalone verifier at /verify/ runs the chain and artifact checks).
- Right of reply
- Authors may submit a response, which is published alongside this report unedited.