-
Add ONNX INT8 model (156MB, no torch dependency needed)
-
Update to v8: density-aware compression (293K Sonnet 4.6 labels with density tags)
-
Update: v9 results (7.9/10 quality), add kompress-small comparison
-
Update model card: v9 evals, architecture details, comparison with LLMLingua-2
-
v9: 6.9/10 quality (beats LLMLingua-2 6.2), 2.3x faster, 215K extractive labels from 8 diverse datasets
-
v4: coherent labels (37K) — Claude-distilled extractive compression, argmax inference
-
v3: trained on 330K structured tool outputs (H100) — JSON, diffs, logs, code, SQL, agentic traces
-
v2: trained on 15K labels (3x more data) from agentic + function-calling + coding traces
-
Initial release: Kompress v1 — ModernBERT token compressor for agentic contexts
-
initial commit
