context-compressor
Public
bench / results / 2026-08-24-qwen-qwen3-1.7b.md
bench / results / 2026-08-24-qwen-qwen3-1.7b.md
Model: qwen/qwen3-1.7b
Settings: {"limitTokens":500,"floorTokens":400,"keepRecentTokens":300,"chunkTokens":800,"summarizerMaxTokens":400,"consolidationBudgetTokens":600,"consolidationKeepNewest":2,"mergeMaxTokens":400,"maxExcerptChars":32000,"quizMaxTokens":120}
| Fixture | Recall (compacted) | Recall (consolidated) | Reduction | Chunk latency mean/p50/max (ms) | Structure | Cache |
|---|---|---|---|---|---|---|
| coding-agent | 2/5 | — | 64% | 398/304/670 | ✅ | ✅ |
| research | 3/3 | — | 42% | 750/750/750 | ✅ | ✅ |
| contradictory-updates | 3/3 | — | 31% | 419/419/419 | ✅ | ✅ |
| tool-spam | 3/3 | 3/3 | 53% | 779/661/1370 | ✅ | ✅ |
Recall = seeded facts answered correctly from the compacted view alone. Results are comparable only within one model; see bench/README.md.
Model: qwen/qwen3-1.7b
Settings: {"limitTokens":500,"floorTokens":400,"keepRecentTokens":300,"chunkTokens":800,"summarizerMaxTokens":400,"consolidationBudgetTokens":600,"consolidationKeepNewest":2,"mergeMaxTokens":400,"maxExcerptChars":32000,"quizMaxTokens":120}
| Fixture | Recall (compacted) | Recall (consolidated) | Reduction | Chunk latency mean/p50/max (ms) | Structure | Cache |
|---|---|---|---|---|---|---|
| coding-agent | 2/5 | — | 64% | 398/304/670 | ✅ | ✅ |
| research | 3/3 | — | 42% | 750/750/750 | ✅ | ✅ |
| contradictory-updates | 3/3 | — | 31% | 419/419/419 | ✅ | ✅ |
| tool-spam | 3/3 | 3/3 | 53% | 779/661/1370 | ✅ | ✅ |
Recall = seeded facts answered correctly from the compacted view alone. Results are comparable only within one model; see bench/README.md.