|
Download docs/experiments.md from haidang2405/telen: direct link, hf CLI and curl.
- Browser
- Download file 2.15 kB
-
https://huggingface.co/haidang2405/telen/resolve/main/docs/experiments.md
- Command line
-
hf download hf://haidang2405/telen/docs/experiments.md
-
curl -L -o experiments.md https://huggingface.co/haidang2405/telen/resolve/main/docs/experiments.md
2.15 kB
Experiment commands
The scripts in this repository reproduce data preparation, controls, and evaluation logic. Datasets, fitted checkpoints, embeddings, ranking files, and result receipts are excluded from the public release.
Core mechanism
Run the local update contract without external data:
python examples/frozen_update_demo.py
python -m pytest -q tests/test_index_and_update.py tests/test_scorer_and_training.py
Static retrieval
scripts/run_static_benchmark.pyprepares BM25, dense, fusion, and reranker comparisons.scripts/run_alqac_benchmark.pyloads organizer-provided ALQAC files from a local path.scripts/run_alqac_evaluation.pyevaluates the TELEN static/no-event path when compatible temporal evidence is unavailable.
The static scripts include optional compatibility support for the predecessor text encoder. They do not load predecessor weights unless the caller supplies a local checkpoint.
Learned graph controls
scripts/run_drill_graph_evaluation.pybuilds the shared candidate and text-score inputs.scripts/run_drill_strict_graph_residual_oof.pyruns the strict typed graph residual under out-of-fold splits.scripts/analyze_drill_transparent_graph_control.pyevaluates transparent propagation controls.scripts/run_drill_lora_cross_encoder_oof.pyevaluates the separately tuned text control.scripts/run_drill_to_alqac_graph_transfer.pymeasures graph transfer without fitting on ALQAC labels.
Temporal checks
scripts/build_temporal_pilot.pyconstructs legal-event inputs from locally obtained sources.scripts/run_ai_silver_temporal_experiment.pyruns the development-only intervention protocol.scripts/run_fiscalqa_external_validation.pyevaluates transparent temporal version resolution.scripts/run_fiscalqa_residual_oof.pymeasures whether a learned residual adds to that resolver.
Release audit
Before publishing a source snapshot, run:
python scripts/audit_public_release.py
The audit rejects private filenames, benchmark records, fitted weights, embedding caches, ranking outputs, and other restricted artifacts.