Washington-state-law-embedding-model-Large

Washington-state-law-embedding-model-Large is a highly specialized, parameter-rich embedding model fine-tuned specifically for Legal Information Retrieval (IR) within the State of Washington.

Generic embedding models often perform suboptimally on legal texts due to the semantic gap between natural language questions (e.g., "What dollar amount makes a theft a first degree felony?") and formal statutory legalese. This model bridges that gap, allowing plain-English queries, legal scenarios, and document drafts to be accurately mapped to their corresponding Washington State statutes (Revised Code of Washington - RCW).

Available Models

Model	Language	Description	Query Prefix
CSI-lab/Washington-state-law-embedding-model-Large	English	Fine-tuned large model (1024d) for WA State RCWs. Best performance.	Represent this sentence for searching relevant passages:
CSI-lab/Washington-state-law-embedding-model-Base	English	Fine-tuned base model (768d) for WA State RCWs. Faster inference.	Represent this sentence for searching relevant passages:

Model Overview

• Base Model: BAAI/bge-large-en-v1.5
• Task: Semantic Search / Information Retrieval / Legal Preemption Analysis
• Language: English (Legal Domain)
• Max Sequence Length: 512 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity

Key Features

• Fine-tuned for Washington State legal domain (RCW)
• Optimized for semantic search and retrieval tasks
• Supports natural language legal queries
• Designed for RAG-based legal assistants
• Superior retrieval capacity leveraging the 1024d large architecture

Intended Use Cases

This model is optimized to act as the retriever component in legal Retrieval-Augmented Generation (RAG) pipelines. Primary use cases include:

1. Statutory Cross-Referencing: Mapping natural language legal questions to specific RCWs.
2. Preemption Checking: Automatically retrieving state laws that may preempt or conflict with proposed municipal ordinances.
3. Legal Research Automation: Clustering and searching local agency drafts against established state frameworks.
4. AI Legal Assistants: Powering chatbots and research tools that require accurate retrieval of Washington State laws before generating an answer.
5. Automated Compliance: Scanning contracts or external drafts against established state legislative frameworks.

Technical Details & Training Methodology

The Semantic Gap

A standard dense retriever often fails on legal tasks because it relies on vocabulary overlap rather than conceptual legal mapping. To address this, Washington-state-law-embedding-model-Large was fine-tuned using a synthetic, high-variance dataset.

Training Data

The model was fine-tuned on synthetic legal query–passage pairs generated from Washington State RCW statutes.

The dataset includes:

• Size: 455,424 training samples
• Natural language paraphrases of legal questions
• Hypothetical legal scenarios
• Statute-grounded positive document matches

The dataset spans 500+ legal categories derived from RCW structure.

Hyperparameters & Architecture

• Loss Function: Multiple Negatives Ranking (MNR) Loss
• Batch Size: 32
• Epochs: 4
• fp16: True
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: round_robin
• Learning Rate Decay: Linear
• Infrastructure: High-Performance Computing (HPC) Cluster

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1
• num_train_epochs: 4
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: True
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 42
• data_seed: None
• jit_mode_eval: False
• use_ipex: False
• bf16: False
• fp16: True
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_min_num_params: 0
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• fsdp_transformer_layer_cls_to_wrap: None
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• parallelism_config: None
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: None
• hub_always_push: False
• hub_revision: None
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• include_for_metrics: []
• eval_do_concat_batches: True
• fp16_backend: auto
• push_to_hub_model_id: None
• push_to_hub_organization: None
• mp_parameters:
• auto_find_batch_size: False
• full_determinism: False
• torchdynamo: None
• ray_scope: last
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

Evaluation Metrics

The model was evaluated on a rigorously held-out validation set of synthetic municipal drafts mapped 1-to-1 against Washington State RCWs. The table below compares the peak validation performance (achieved at Epoch 3.02) against the baseline, untrained bge-large model.

Metric	Base Model (Untrained Large)	Fine-Tuned (Peak @ 3.02)	Absolute Improvement
Recall@10	0.5684	0.8354	+ 26.7%
Recall@5	0.2842	0.4255	+ 14.13%
NDCG@10	0.2509	0.3828	+ 12.38%
MRR@10	0.1569	0.2487	+ 9.18%

Interpretation: Because the BAAI large architecture is already highly proficient, the baseline was extremely strong out-of-the-box. Fine-tuning pushed the model to extract the absolute mathematical ceiling from this legal dataset, successfully returning the exact governing state law within the top 10 results 83.5% of the time.

Limitations

• This model does not provide legal advice.
• Performance is limited to Washington State law (RCW) and may not generalize to other jurisdictions.
• Outputs depend on the quality of the underlying document corpus.
• Should be used as a retrieval tool, not a final decision-making system.

Usage Examples

Semantic Search with sentence-transformers

Warning: Because this model is built on the BGE architecture, you must append the specific instruction prefix
"Represent this sentence for searching relevant passages:"
to your search queries to achieve optimal performance.

Do not add this prefix to the database documents.

import torch
from sentence_transformers import SentenceTransformer, util

# 1. Load the fine-tuned model
model = SentenceTransformer('CSI-lab/Washington-state-law-embedding-model-Large')

# 2. Define the laws (Your Vector Database)
laws = [
    "RCW 9A.56.030: Theft in the first degree. A person is guilty of theft in the first degree if he or she commits theft of property or services which exceed(s) five thousand dollars in value.",
    "RCW 46.61.502: Driving under the influence. A person is guilty of driving while under the influence of intoxicating liquor...",
    "RCW 9A.36.011: Assault in the first degree. A person is guilty of assault in the first degree if he or she..."
]

# 3. Define the user's search query
user_query = "What dollar amount makes a theft a first degree felony?"

# 4. CRITICAL: Add the required BGE prefix to the query ONLY
query_prefix = "Represent this sentence for searching relevant passages: "
formatted_query = query_prefix + user_query

# 5. Encode the documents and the query
law_embeddings = model.encode(laws, convert_to_tensor=True)
query_embedding = model.encode(formatted_query, convert_to_tensor=True)

# 6. Calculate Cosine Similarity
cosine_scores = util.cos_sim(query_embedding, law_embeddings)

# 7. Print the top result
best_idx = cosine_scores.argmax().item()
print(f"Top Match: {laws[best_idx]}")
print(f"Similarity Score: {cosine_scores[0][best_idx]:.4f}")

Model Citation

@misc{washington_state_law_embedding_Large_2026,
  title={Washington-state-law-embedding-model-Large: Fine-Tuned Dense Retrieval for Washington State Law},
  author={Tomar, Shlok},
  year={2026},
  publisher={Hugging Face}
  howpublished={\url{https://huggingface.co/CSI-lab/Washington-state-law-embedding-model-Large}},
  note={Hugging Face Model Repository}
}

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}