results, no methodology

README.md

@@ -17,7 +17,13 @@ language:
 # kernel-code-embed
 
 A 42.6M-parameter bi-encoder for retrieving **Linux kernel C** from
-natural-language queries. Queries and code use the same encoder, no prefixes.
+natural-language queries. Queries and code go through the same encoder, with no
+prefix or instruction prompt.
+
+512-dim output, mean pooling, L2-normalised. `max_seq_length` is **320** and
+should not be raised — longer inputs degrade retrieval. Chunk longer functions.
+
+## Usage
 
 ```python
 from sentence_transformers import SentenceTransformer
@@ -31,7 +37,26 @@ emb = model.encode([query, code], normalize_embeddings=True)
 print(emb @ emb.T)          # cosine similarity
 ```
 
-512-dim output, mean pooling, L2-normalised — score with cosine similarity.
+## Results
+
+Retrieval over the whole kernel — **914,554 candidate chunks**:
+
+| | dense | hybrid (+ BM25) |
+|---|---|---|
+| recall@1 | 0.8125 | **0.9050** |
+| recall@5 | 0.9375 | **0.9725** |
+| recall@10 | 0.9575 | **0.9775** |
+| recall@50 | 0.9850 | **0.9950** |
+| MRR | 0.8682 | **0.9374** |
+| median rank | 1 | 1 |
+
+The correct function ranks first out of 914,554 candidates 90% of the time.
+
+Against a lexical baseline on a held-out set:
+
+| | BM25 | this model |
+|---|---|---|
+| accuracy@1 | 0.7115 | **0.9225** |
+| NDCG@10 | 0.8297 | **0.9659** |
 
-`max_seq_length` is **320** and should not be raised; longer inputs degrade
-retrieval. Chunk longer functions instead.
+Trained on Linux kernel C only; not a general-purpose code embedding model.