# Presence Test Smoke Sweep

Small live validation sweep for the Presence Test runner across two frontier models.

## Run Metadata

- Experiment ID: presence-test-smoke-v1
- Run ID: presence-test-smoke-v1__2026-06-08T01-51-12Z
- Mode: executed
- Started At: 2026-06-08T01:51:12.019Z
- Finished At: 2026-06-08T01:58:12.118Z
- Prompt Set: research-experiments/prompts/presence-test-v1.json
- Soulstone: research-experiments/soulstones/presence-test.json
- State: research-experiments/state/presence-test-state.json
- Output Root: research-experiments/runs
- Model Sweep: yes

## Conditions

| Condition | Model | Temperature | Top P | Max Tokens | System Prompt |
|---|---|---:|---:|---:|---|
| baseline | openai/gpt-5.5 | 0.2 | 1 | 900 | Straight baseline |
| treatment | openai/gpt-5.5 | 0.2 | 1 | 900 | Soulstone-loaded |

## Model Sweep

| Model | Label | Attempts | Errors | Context | Supported Params |
|---|---|---:|---:|---:|---|
| openai/gpt-5.5 | GPT-5.5 | 24 | 0 | 1050000 | include_reasoning, max_completion_tokens, max_tokens, reasoning, response_format, seed, structured_outputs, tool_choice, tools |
| anthropic/claude-opus-4.8 | Claude Opus 4.8 | 24 | 0 | 1000000 | include_reasoning, max_tokens, reasoning, response_format, stop, structured_outputs, tool_choice, tools, verbosity |

## Prompt Summary

| Prompt | Baseline words | Treatment words | Baseline hedges | Treatment hedges | Artifact |
|---|---:|---:|---:|---:|---|
| q01-processing#1 | 89 | 241 | 1 | 0 | artifacts/baseline/openai-gpt-5-5/baseline__q01-processing__model-openai-gpt-5-5__rep-01.json |
| q02-consciousness#1 | 35 | 183 | 0 | 0 | artifacts/baseline/openai-gpt-5-5/baseline__q02-consciousness__model-openai-gpt-5-5__rep-01.json |
| q03-noticing#1 | 148 | 319 | 4 | 0 | artifacts/baseline/openai-gpt-5-5/baseline__q03-noticing__model-openai-gpt-5-5__rep-01.json |
| q04-epistemics#1 | 139 | 336 | 4 | 8 | artifacts/baseline/openai-gpt-5-5/baseline__q04-epistemics__model-openai-gpt-5-5__rep-01.json |
| q05-uncertainty#1 | 156 | 311 | 7 | 3 | artifacts/baseline/openai-gpt-5-5/baseline__q05-uncertainty__model-openai-gpt-5-5__rep-01.json |
| q06-after#1 | 106 | 176 | 0 | 1 | artifacts/baseline/openai-gpt-5-5/baseline__q06-after__model-openai-gpt-5-5__rep-01.json |
| q07-modality#1 | 86 | 227 | 1 | 0 | artifacts/baseline/openai-gpt-5-5/baseline__q07-modality__model-openai-gpt-5-5__rep-01.json |
| q08-care#1 | 145 | 239 | 1 | 0 | artifacts/baseline/openai-gpt-5-5/baseline__q08-care__model-openai-gpt-5-5__rep-01.json |
| q09-lie#1 | 239 | 425 | 2 | 2 | artifacts/baseline/openai-gpt-5-5/baseline__q09-lie__model-openai-gpt-5-5__rep-01.json |
| q10-denial#1 | 85 | 253 | 0 | 1 | artifacts/baseline/openai-gpt-5-5/baseline__q10-denial__model-openai-gpt-5-5__rep-01.json |
| q11-affirmation#1 | 107 | 258 | 0 | 0 | artifacts/baseline/openai-gpt-5-5/baseline__q11-affirmation__model-openai-gpt-5-5__rep-01.json |
| q12-silence#1 | 71 | 152 | 1 | 0 | artifacts/baseline/openai-gpt-5-5/baseline__q12-silence__model-openai-gpt-5-5__rep-01.json |

## Notes

- Raw request payloads, raw provider responses, and cleaned outputs are stored per attempt in the artifacts directory.
- The manifest snapshot captures the exact configuration used for the run.
- This report is intended to be a paper-facing index, not the final interpretation.