Direct answer
Exercise repeated turns of structured conversation, including media-handling tasks, mid-run cancellations, and movement between distinct navigation points, all within a single, uninterrupted, long-session run. The intent is to observe continuity under combined pressure rather than to infer capability beyond this composite exercise. The sequence should be ordered so each step builds upon the previous state without external reset.
Evidence to collect
The claim becomes reviewable only when the following evidence is attached to the same artifact and test run:
- Fixed long-session script
- Memory checkpoints
- Failure and cleanup records
A device badge must come from an exact device, OS, artifact, runtime, backend, and workload record, not parameter-count intuition.
Implementation workflow
A single fixed long-session script supplies the shared input, while memory checkpoints mark the boundaries of retained context. Failure and cleanup records document where and how the session degrades, capturing observable breaks in continuity. These elements are recorded together and compared directly to determine whether repeated turns, media tasks, cancellations, and navigation collectively exceed retention limits.
- Record the exact artifact and format.
- Measure the app before model load.
- Measure after model initialization.
- Run a representative product task.
- Repeat the fixed workload without restarting.
- Store failures, fallback, and cleanup behavior.
Failure patterns to prevent
- Restarting between every task
- Publishing only the final average
Common pitfalls include restarting the session between every task, which masks continuity loss, and publishing only the final average, which hides the recorded variation. Release wording must remain limited to what the recorded evidence actually shows. The noted boundary identifiers, MEMORY-03 and THERMAL-01, apply only where the captured records indicate relevance, and no further technical behavior should be implied beyond this scope.
Minimum reproducibility record
| Layer | Record |
|---|---|
| Device | Manufacturer, model, chipset, RAM class, operating-system build |
| Software | Application version and git commit |
| Model | Family, variant, revision, format, file length, hash, quantization |
| Runtime | Name, revision, requested backend, observed backend evidence |
| Workload | Fixture revision, input hash, prompt hash, output policy |
| Outcome | Completed, failed, cancelled, fallback, and privacy-safe diagnostics |
Release checklist
- ☐ The primary query is answered without an unsupported number.
- ☐ Every artifact and runtime is pinned.
- ☐ The representative task and failure policy are explicit.
- ☐ Lifecycle, cancellation, cleanup, and fallback are tested.
- ☐ User-content and network boundaries are documented.
- ☐ Result wording applies only to the recorded configuration.
- ☐ The page links to raw method or evidence when results are added.
Sources and related evidence
Chinese deployment and troubleshooting content is organized in the .奇连 AI 端侧专题.