Downloads · 30 days
35
7% of all-time downloads
Steve0927/qwen3-embedding-0.6b-lrat-full80k-hardneg-preserve-mr2-min6-from-broad
qwen3-embedding-0.6b-lrat-full80k-hardneg-preserve-mr2-min6-from-broad is a sentence similarity model from Steve0927. Use it when you need a score for how close two texts are. It is set up for sentence-transformers.
This is a Qwen3-Embedding-0.6B compatible dense retriever checkpoint prepared for XIR competition 1170.
Downloads · 30 days
35
7% of all-time downloads
All-time downloads
481
Public
Parameters
596M
1.2 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors1.2 GB · 99%
From the Hugging Face model README
This is a Qwen3-Embedding-0.6B compatible dense retriever checkpoint prepared for XIR competition 1170.
The method is Trajectory Utility Calibration (TUC).
We train only from official LRAT trajectory-derived retrieval rows. LRAT converts agent search/browse/reasoning trajectories into query-positive-negative training examples. TUC adds a conservative second-pass row-internal audit:
Training data source:
Base checkpoint:
Compliance boundary:
3e-70.51624drop_low_pos_quality<=1.0; false_negative relevance>=3.5 confidence>=0.75; max_remove_per_row=2; min_neg=6; weight=[0.4,2.0]Full80k gate assessment:
passreports/full80k_gate_assessment.jsonDense gate:
reports/dense_gate_full80k_hardneg_preserve_mr2_min6_vs_full80k_conservative.jsonpass vs broad_g16_v2 and pass vs full80k_conservative; recall@50 +0.001185 vs current best; hit@50 flatThis checkpoint is intended as a dense retriever in the XIR/BrowseComp-Plus-style fixed-corpus agent-search setting, where the submitted artifact is only the retriever checkpoint and the agent/judge/runtime are fixed externally.