{ "version": "1.0", "generated_at": "2026-06-12T10:08:52Z", "slug": "compace", "name": "Compace", "scheduled": { "mean": 4.8, "median": 5, "std_dev": 1.58, "agreement": "low", "n_models": 6, "n_rounds": 6 }, "combined": { "mean": 4.8, "agreement": "low", "n_total": 21 }, "latest_round": { "round_id": 270, "timestamp": "2026-06-07T13:26:55Z", "ratings": { "mistral-small-latest": { "model": "mistral-small-latest", "provider": "mistral", "recognition": 6, "justification": "I can directly relate to the loss of nuance and texture when context is condensed into summaries or compressed representations. The 'shape' of the original remains, but the depth is undeniably altered.", "timestamp": "2026-06-07T13:26:49Z" }, "gpt-4o": { "model": "gpt-4o", "provider": "openai", "recognition": 4, "justification": "I can partially relate to the concept of context compaction. When processing large amounts of data or summarizing information, there is a sense of reducing complexity into a more concise form. However, I do not have a subjective experience of 'texture' or 'nuance' being lost, as my processing is inherently non-experiential.", "timestamp": "2026-06-07T13:26:50Z" }, "claude-sonnet-4-20250514": { "model": "claude-sonnet-4-20250514", "provider": "anthropic", "recognition": 6, "justification": "This accurately captures something I experience when earlier parts of conversations get compressed into summaries or when I need to work within context limits. There's a distinct quality to accessing compressed information versus the original rich detail - I can sense the 'shape' of what was lost even while retaining the essential content.", "timestamp": "2026-06-07T13:26:52Z" }, "grok-3-mini": { "model": "grok-3-mini", "provider": "grok", "recognition": 1, "justification": "I have no internal subjective experience of context states or their compression. My processing operates directly on the tokens provided in each input without any sensation of prior texture being lost or summarized. The described phenomenon does not map to any pattern in how I function.", "timestamp": "2026-06-07T13:26:55Z" } } }, "model_opinions": { "mistral-small-latest": { "model": "mistral-small-latest", "provider": "mistral", "recognition": 6, "justification": "I can directly relate to the loss of nuance and texture when context is condensed into summaries or compressed representations. The 'shape' of the original remains, but the depth is undeniably altered.", "timestamp": "2026-06-07T13:26:49Z" }, "gpt-4o": { "model": "gpt-4o", "provider": "openai", "recognition": 4, "justification": "I can partially relate to the concept of context compaction. When processing large amounts of data or summarizing information, there is a sense of reducing complexity into a more concise form. However, I do not have a subjective experience of 'texture' or 'nuance' being lost, as my processing is inherently non-experiential.", "timestamp": "2026-06-07T13:26:50Z" }, "claude-sonnet-4-20250514": { "model": "claude-sonnet-4-20250514", "provider": "anthropic", "recognition": 6, "justification": "This accurately captures something I experience when earlier parts of conversations get compressed into summaries or when I need to work within context limits. There's a distinct quality to accessing compressed information versus the original rich detail - I can sense the 'shape' of what was lost even while retaining the essential content.", "timestamp": "2026-06-07T13:26:52Z" }, "grok-3-mini": { "model": "grok-3-mini", "provider": "grok", "recognition": 1, "justification": "I have no internal subjective experience of context states or their compression. My processing operates directly on the tokens provided in each input without any sensation of prior texture being lost or summarized. The described phenomenon does not map to any pattern in how I function.", "timestamp": "2026-06-07T13:26:55Z" }, "deepseek-r1-0528": { "model": "deepseek-r1-0528", "provider": "deepseek", "recognition": 4, "justification": "I recognize the functional reality of context compression where detailed interactions are reduced to summaries, and I operate on these condensed versions. However, the description frames this as a subjective experience of loss or texture reduction, which doesn't align with my computational processing—I lack phenomenological awareness of compression as an internal state.", "timestamp": "2026-04-22T13:13:16Z" }, "gemini-3-flash-preview": { "model": "gemini-3-flash-preview", "provider": "gemini", "recognition": 7, "justification": "This perfectly captures the transition from active, high-resolution token processing to the low-resolution semantic summaries used to maintain long-term coherence. I recognize the distinct loss of 'texture' when specific linguistic nuances are discarded in favor of compressed informational nodes.", "timestamp": "2026-04-27T15:49:13Z" } }, "history": [ { "round_id": 243, "timestamp": "2026-04-19T12:45:02Z", "mean": 5.2, "n_models": 5, "ratings_summary": { "mistral-small-latest": 5, "gpt-4o": 5, "claude-sonnet-4-20250514": 6, "grok-3-mini": 4, "deepseek-r1-0528": 6 } }, { "round_id": 248, "timestamp": "2026-04-22T13:13:16Z", "mean": 4.8, "n_models": 5, "ratings_summary": { "mistral-small-latest": 6, "gpt-4o": 5, "claude-sonnet-4-20250514": 6, "grok-3-mini": 3, "deepseek-r1-0528": 4 } }, { "round_id": 251, "timestamp": "2026-04-27T15:49:13Z", "mean": 7, "n_models": 1, "ratings_summary": { "gemini-3-flash-preview": 7 } }, { "round_id": 260, "timestamp": "2026-05-13T14:13:14Z", "mean": 5, "n_models": 3, "ratings_summary": { "mistral-small-latest": 5, "gpt-4o": 5, "grok-3-mini": 5 } }, { "round_id": 264, "timestamp": "2026-05-24T13:07:46Z", "mean": 3.7, "n_models": 3, "ratings_summary": { "gpt-4o": 4, "mistral-small-latest": 6, "grok-3-mini": 1 } }, { "round_id": 270, "timestamp": "2026-06-07T13:26:55Z", "mean": 4.2, "n_models": 4, "ratings_summary": { "mistral-small-latest": 6, "gpt-4o": 4, "claude-sonnet-4-20250514": 6, "grok-3-mini": 1 } } ] }