Observed since 20 Oct, 18:46 UTC
No superseded metadata states collected. Unchanged observations do not create new revisions.
AdaDec is an uncertainty-guided adaptive decoding framework for large language model (LLM) code generation which enhances standard decoding with a selective pause-then-rerank mechanism triggered by model uncertainty.
AdaDec is an adaptive decoding framework for LLM code generation. It selectively pauses decoding when uncertainty is high, reranks candidates, and improves accuracy with moderate overhead.
pip install adadec
from adadec import prepare_adadec, generate_adadec
prepare_adadec(model, tokenizer, "train.jsonl", "out.parquet", "mymodel", "thresholds.json")
result = generate_adadec(model, tokenizer, ["def add(a, b):"], "mymodel")
Note that train.jsonl requires "task_id", "prompt", and "canonical_solution".
stop_words.json example:
["\\n{4,}", "^\\S"]
MIT License
Download counts and popularity will appear here as the collector receives them.
| Version | Published | Direct dependencies | Open version |
|---|---|---|---|
| 0.1.0 | 25 Sept 2025 | 5 |
This describes our collected data; it is not a security assessment.
Observed since 20 Oct, 18:46 UTC
No superseded metadata states collected. Unchanged observations do not create new revisions.
package:d0b5cc4aaa956f9aed6c1952c3d327684c77dd3f4afb794c86193d4ae6f4493a