eric-ml-nlp/Delveta-LayaChoice-v1-data

Delveta LayaChoice v1 — frozen dataset

Purpose

This dataset is the frozen evaluation and training corpus for Delveta-LayaChoice-v1, a capability-selection decision model used inside Delveta's chat Intent Funnel. It exists so that the fine-tuning result is auditable and reproducible: the exact bytes used for training, validation and the final benchmark are published here with their SHA256 digests.

Task

Each row is a capability-selection (multiple-choice) example:

• one user turn (the "state") in Chinese or English;
• three candidate capability cards (K = 3), rendered from the B_noprov view;
• a single gold capability index.

The model must choose the gold capability. Cards come from a fixed roster of 17 capabilities:

cap-add-term          cap-artifact          cap-bash
cap-create-folder     cap-edit-file         cap-mindmap
cap-pdf-extract-text  cap-pdf-table-to-text cap-rag-search
cap-read-document     cap-read-file         cap-slides
cap-social-search     cap-summary           cap-translate
cap-vision            cap-web-search

Splits

Split	Rows	File	Role
train	856	data/layachoice_v1_train_b_noprov.jsonl.gz	source training corpus
validation	150	data/layachoice_v1_val_b_noprov.jsonl.gz	checkpoint selection only
test	900	data/layachoice_v1_test_b_noprov.jsonl.gz	final benchmark only

The training corpus of 856 rows is split further at model-training time: 85 rows are held out for temperature calibration (fixed IDs, seed 20260922), leaving 771 effective training rows. Those 85 calibration rows are part of the 856 published here; they were never seen by the optimizer.

raw/ holds the two non-regenerable source artifacts the bundles were rendered from:

• raw/LayaChoice_v1_validation_raw_v3.jsonl
• raw/LayaChoice_v1_final_test_raw_v3.jsonl

Frozen status

The dataset is frozen. Any change to a file here invalidates the published benchmark. Verify with:

sha256sum -c SHA256SUMS

File	SHA256
data/layachoice_v1_train_b_noprov.jsonl.gz	e968fdeddc8f02eec19f37ed29fafc5f05a591ea259c97315639d366e3de4f13
data/layachoice_v1_val_b_noprov.jsonl.gz	3fa38aa099580b6f9a5146ac2a8e7beed0111d79d2f994adb6ca317547c90364
data/layachoice_v1_test_b_noprov.jsonl.gz	c665004a127496f4456a0b3ccd81bd0ce10654fef7808f27e45500d3a9d9b8bb
raw/LayaChoice_v1_validation_raw_v3.jsonl	974a3b99cf5883c527677373b53539f3ac859ddfb0ba59ea793058d03a6710ff
raw/LayaChoice_v1_final_test_raw_v3.jsonl	e0738991d193cae39531a384305dc5d3547ffdd69f9e80b062c91079f51b4613

Source-file digests (the artifacts the bundles were rendered from) are recorded in the model's experiment document.

Row schema

Each line of a bundle is a compact JSON record:

{"id": "...", "lang": "zh|en", "state": "...",
 "order": ["cap-a", "cap-b", "cap-c"],
 "options": ["<card text>", "<card text>", "<card text>"],
 "gold": 0, "d1": 1, "d2": 2, "src_sha": "..."}

• order is the candidate capability order presented to the model;
• options are the three B_noprov card texts in order;
• gold is the index into order of the gold capability;
• d1 / d2 index the first and second hard-negative distractors (used for the gold→d1 / gold→d2 error breakdown, not for training).

Data construction notes

• Cards are rendered from the B_noprov view: a capability's description without any evidence: line or provenance that would reveal the label. The same view is used for training, validation and the benchmark, so comparisons are like-for-like.
• Exclusion / leakage prevention. A row's card text is asserted not to contain the user turn verbatim; ingest hard-fails with an ExclusionSetError if the check is violated. Train / validation / test are disjoint by construction, and the 85 calibration rows are held out of the optimizer by explicit ID.
• The corpus is synthetic / curated — it was authored for this task, not sampled from live product traffic. Hard negatives (d1, d2) are chosen to be semantically near the gold capability, which is the intended difficulty of the benchmark.

Known benchmark limitations

• The test split is a fixed, curated 900-row artifact over 17 known capabilities with a known hard-negative construction. Accuracy on it is benchmark accuracy — not production-traffic accuracy and not open-world generalization accuracy.
• The candidate count is fixed at K = 3. Behaviour for K = 2 or post-truncation K >= 4 is not represented.
• Capabilities introduced after the freeze are out of scope.
• Because the corpus is synthetic, it carries no distributional guarantee with respect to live user requests.

Licensing status

No license is asserted for this dataset. The laya software package is Apache-2.0, but the license of the base model weights (convaiinnovations/laya) could not be determined at release time. This repository is private; downstream use should confirm terms before redistribution.

Reference

Full method, ablations, baselines and invalid runs: docs/experiments/LayaChoice-v1-Fine-Tuning.md.