model_id: complexity-deberta-v3-train-pool
domain: complexity
display_name: DeBERTa V3 train_pool
description: DeBERTa V3 sequence classifier for CEFR-based text complexity classification.
version: '2026-03-07'
is_active: false
priority: 1
framework:
  type: transformers
  task: sequence-classification
  library: huggingface
  backbone: microsoft/deberta-v3-base
  architecture: DebertaV2ForSequenceClassification
artifacts:
  weights:
  - model.safetensors
  tokenizer:
  - tokenizer.json
  - tokenizer_config.json
  - spm.model
  - special_tokens_map.json
  - added_tokens.json
  config:
  - config.json
  vocabulary: []
model: {}
runtime:
  max_sequence_length: 256
  truncation: true
  padding: true
  batch_size: 8
  device: auto
inference:
  return_probabilities: true
  confidence_method: softmax_max
ui:
  domain_display_name: Text Complexity
  color_token: complexity
  group: complexity-core
labels:
  type: single-label-classification
  classes:
  - id: 0
    name: A1-A2
    display_name: A1-A2
  - id: 1
    name: B1-B2
    display_name: B1-B2
  - id: 2
    name: C1-C2
    display_name: C1-C2