Downloads · 30 days
0
haidang2405/telen
telen is a machine learning model from haidang2405. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for pytorch. The card lists the license as mit.
[](https://github.com/dangdinh2405/telen) [](https://github.com/dangdinh2405/telen/blob/master/LICENSE)
Downloads · 30 days
0
Access
Public
Updated Aug 30, 2026
Repo size
—
Likes
0
Public
Click a slice to open those files.
.py1.1 MB · 99%
From the Hugging Face model README
TELEN is a query-conditioned graph residual system for Vietnamese statutory article retrieval. It stores provision vectors in a stable text space, appends new legal versions without re-encoding older vectors, and adds a bounded local residual only when the candidate graph exposes a typed edge.
This Hub repository mirrors the tested source release. It contains no manuscript, benchmark records, trained checkpoint, embedding cache, or per-query rankings. The absence of model weights is intentional.
legal update
-> validate new versions and events
-> encode new provision texts only
-> append vectors and graph state
-> verify unchanged parameters and older vectors
query
-> frozen text candidates
-> typed local graph messages
-> structural evidence gate
-> bounded residual
-> final ranking
If a candidate has no exposed graph edge, the residual is exactly zero and its final score equals its frozen text score.
git clone https://github.com/dangdinh2405/telen.git
cd telen
python -m pip install -e ".[dev]"
python examples/frozen_update_demo.py
python -m pytest -q
python scripts/audit_public_release.py
src/telen/: current TELEN package and TELENEngine public API.src/telen/legacy/: compatibility code used only by predecessor-baseline scripts.scripts/: experiment preparation, interventions, controls, and evaluation commands.tests/: offline unit and contract tests.docs/: architecture and experiment guidance.No fitted TELEN checkpoint is distributed. The earlier TELEN-GP checkpoint and all paper materials are held in private archives and are not part of this release.
DRiLL, ALQAC, FiscalQA, and licensed Vietnamese retrieval data must be obtained independently from their providers.