--- language: - en license: apache-2.0 tags: - sentence-transformers - cross-encoder - generated_from_trainer - dataset_size:578402 - loss:BinaryCrossEntropyLoss base_model: answerdotai/ModernBERT-base pipeline_tag: text-ranking library_name: sentence-transformers metrics: - map - mrr@10 - ndcg@10 model-index: - name: ModernBERT-base trained on GooAQ results: - task: type: cross-encoder-reranking name: Cross Encoder Reranking dataset: name: gooaq dev type: gooaq-dev metrics: - type: map value: 0.7285 name: Map - type: mrr@10 value: 0.727 name: Mrr@10 - type: ndcg@10 value: 0.77 name: Ndcg@10 - task: type: cross-encoder-reranking name: Cross Encoder Reranking dataset: name: NanoMSMARCO R100 type: NanoMSMARCO_R100 metrics: - type: map value: 0.4718 name: Map - type: mrr@10 value: 0.4647 name: Mrr@10 - type: ndcg@10 value: 0.5533 name: Ndcg@10 - task: type: cross-encoder-reranking name: Cross Encoder Reranking dataset: name: NanoNFCorpus R100 type: NanoNFCorpus_R100 metrics: - type: map value: 0.3424 name: Map - type: mrr@10 value: 0.5554 name: Mrr@10 - type: ndcg@10 value: 0.3784 name: Ndcg@10 - task: type: cross-encoder-reranking name: Cross Encoder Reranking dataset: name: NanoNQ R100 type: NanoNQ_R100 metrics: - type: map value: 0.5178 name: Map - type: mrr@10 value: 0.5159 name: Mrr@10 - type: ndcg@10 value: 0.5882 name: Ndcg@10 - task: type: cross-encoder-nano-beir name: Cross Encoder Nano BEIR dataset: name: NanoBEIR R100 mean type: NanoBEIR_R100_mean metrics: - type: map value: 0.444 name: Map - type: mrr@10 value: 0.512 name: Mrr@10 - type: ndcg@10 value: 0.5066 name: Ndcg@10 --- # ModernBERT-base trained on GooAQ This is a [Cross Encoder](https://www.sbert.net/docs/cross_encoder/usage/usage.html) model finetuned from [answerdotai/ModernBERT-base](https://huggingface.co/answerdotai/ModernBERT-base) using the [sentence-transformers](https://www.SBERT.net) library. It computes scores for pairs of texts, which can be used for text reranking and semantic search. ## Model Details ### Model Description - **Model Type:** Cross Encoder - **Base model:** [answerdotai/ModernBERT-base](https://huggingface.co/answerdotai/ModernBERT-base) - **Maximum Sequence Length:** 8192 tokens - **Number of Output Labels:** 1 label - **Language:** en - **License:** apache-2.0 ### Model Sources - **Documentation:** [Sentence Transformers Documentation](https://sbert.net) - **Documentation:** [Cross Encoder Documentation](https://www.sbert.net/docs/cross_encoder/usage/usage.html) - **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers) - **Hugging Face:** [Cross Encoders on Hugging Face](https://huggingface.co/models?library=sentence-transformers&other=cross-encoder) ## Usage ### Direct Usage (Sentence Transformers) First install the Sentence Transformers library: ```bash pip install -U sentence-transformers ``` Then you can load this model and run inference. ```python from sentence_transformers import CrossEncoder # Download from the 🤗 Hub model = CrossEncoder("tomaarsen/reranker-ModernBERT-base-gooaq-bce-random") # Get scores for pairs of texts pairs = [ ['is esurance a reputable company?', "Esurance auto insurance earned 4.5 stars out of 5 for overall performance. ... Based on these ratings, Esurance is among NerdWallet's Best Car Insurance Companies for 2020. Esurance offers all the usual coverage options, plus optional coverage including: Emergency roadside assistance."], ['is esurance a reputable company?', 'Coinsurance in property insurance is a means for insurers to obtain rate and premium equality. ... Rates are applied against a specified percentage (100, 90, or 80 percent, for example) of the value to the insured: building, contents, or business income.'], ['is esurance a reputable company?', 'Some employers offer both term life insurance coverage and supplemental life insurance. Term life insurance through your employer generally works like regular term life insurance. ... Supplemental life insurance is similar to a group term life insurance policy, but is typically more limited.'], ['is esurance a reputable company?', "Third party insurance is the legal minimum. This means you're covered if you have an accident causing damage or injury to any other person, vehicle, animal or property. It does not cover any other costs like repair to your own vehicle. You may want to use an insurance broker."], ['is esurance a reputable company?', 'In the United States, corporations have limited liability and the expression corporation is preferred to limited company. A "limited liability company" (LLC) is a different entity. However, some states permit corporations to have the designation Ltd. (instead of the usual Inc.) to signify their corporate status.'], ] scores = model.predict(pairs) print(scores.shape) # (5,) # Or rank different texts based on similarity to a single text ranks = model.rank( 'is esurance a reputable company?', [ "Esurance auto insurance earned 4.5 stars out of 5 for overall performance. ... Based on these ratings, Esurance is among NerdWallet's Best Car Insurance Companies for 2020. Esurance offers all the usual coverage options, plus optional coverage including: Emergency roadside assistance.", 'Coinsurance in property insurance is a means for insurers to obtain rate and premium equality. ... Rates are applied against a specified percentage (100, 90, or 80 percent, for example) of the value to the insured: building, contents, or business income.', 'Some employers offer both term life insurance coverage and supplemental life insurance. Term life insurance through your employer generally works like regular term life insurance. ... Supplemental life insurance is similar to a group term life insurance policy, but is typically more limited.', "Third party insurance is the legal minimum. This means you're covered if you have an accident causing damage or injury to any other person, vehicle, animal or property. It does not cover any other costs like repair to your own vehicle. You may want to use an insurance broker.", 'In the United States, corporations have limited liability and the expression corporation is preferred to limited company. A "limited liability company" (LLC) is a different entity. However, some states permit corporations to have the designation Ltd. (instead of the usual Inc.) to signify their corporate status.', ] ) # [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...] ``` ## Evaluation ### Metrics #### Cross Encoder Reranking * Dataset: `gooaq-dev` * Evaluated with [CrossEncoderRerankingEvaluator](https://sbert.net/docs/package_reference/cross_encoder/evaluation.html#sentence_transformers.cross_encoder.evaluation.CrossEncoderRerankingEvaluator) with these parameters: ```json { "at_k": 10, "always_rerank_positives": false } ``` | Metric | Value | |:------------|:---------------------| | map | 0.7285 (+0.1974) | | mrr@10 | 0.7270 (+0.2030) | | **ndcg@10** | **0.7700 (+0.1787)** | #### Cross Encoder Reranking * Datasets: `NanoMSMARCO_R100`, `NanoNFCorpus_R100` and `NanoNQ_R100` * Evaluated with [CrossEncoderRerankingEvaluator](https://sbert.net/docs/package_reference/cross_encoder/evaluation.html#sentence_transformers.cross_encoder.evaluation.CrossEncoderRerankingEvaluator) with these parameters: ```json { "at_k": 10, "always_rerank_positives": true } ``` | Metric | NanoMSMARCO_R100 | NanoNFCorpus_R100 | NanoNQ_R100 | |:------------|:---------------------|:---------------------|:---------------------| | map | 0.4718 (-0.0178) | 0.3424 (+0.0814) | 0.5178 (+0.0982) | | mrr@10 | 0.4647 (-0.0128) | 0.5554 (+0.0555) | 0.5159 (+0.0892) | | **ndcg@10** | **0.5533 (+0.0129)** | **0.3784 (+0.0534)** | **0.5882 (+0.0875)** | #### Cross Encoder Nano BEIR * Dataset: `NanoBEIR_R100_mean` * Evaluated with [CrossEncoderNanoBEIREvaluator](https://sbert.net/docs/package_reference/cross_encoder/evaluation.html#sentence_transformers.cross_encoder.evaluation.CrossEncoderNanoBEIREvaluator) with these parameters: ```json { "dataset_names": [ "msmarco", "nfcorpus", "nq" ], "rerank_k": 100, "at_k": 10, "always_rerank_positives": true } ``` | Metric | Value | |:------------|:---------------------| | map | 0.4440 (+0.0539) | | mrr@10 | 0.5120 (+0.0440) | | **ndcg@10** | **0.5066 (+0.0513)** | ## Training Details ### Training Dataset #### Unnamed Dataset * Size: 578,402 training samples * Columns: question, answer, and label * Approximate statistics based on the first 1000 samples: | | question | answer | label | |:--------|:-----------------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------------------|:------------------------------------------------| | type | string | string | int | | details | | | | * Samples: | question | answer | label | |:----------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------| | is esurance a reputable company? | Esurance auto insurance earned 4.5 stars out of 5 for overall performance. ... Based on these ratings, Esurance is among NerdWallet's Best Car Insurance Companies for 2020. Esurance offers all the usual coverage options, plus optional coverage including: Emergency roadside assistance. | 1 | | is esurance a reputable company? | Coinsurance in property insurance is a means for insurers to obtain rate and premium equality. ... Rates are applied against a specified percentage (100, 90, or 80 percent, for example) of the value to the insured: building, contents, or business income. | 0 | | is esurance a reputable company? | Some employers offer both term life insurance coverage and supplemental life insurance. Term life insurance through your employer generally works like regular term life insurance. ... Supplemental life insurance is similar to a group term life insurance policy, but is typically more limited. | 0 | * Loss: [BinaryCrossEntropyLoss](https://sbert.net/docs/package_reference/cross_encoder/losses.html#binarycrossentropyloss) with these parameters: ```json { "activation_fct": "torch.nn.modules.linear.Identity", "pos_weight": 5 } ``` ### Training Hyperparameters #### Non-Default Hyperparameters - `eval_strategy`: steps - `per_device_train_batch_size`: 64 - `per_device_eval_batch_size`: 64 - `learning_rate`: 2e-05 - `num_train_epochs`: 1 - `warmup_ratio`: 0.1 - `seed`: 12 - `bf16`: True - `dataloader_num_workers`: 4 - `load_best_model_at_end`: True #### All Hyperparameters
Click to expand - `overwrite_output_dir`: False - `do_predict`: False - `eval_strategy`: steps - `prediction_loss_only`: True - `per_device_train_batch_size`: 64 - `per_device_eval_batch_size`: 64 - `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`: 2e-05 - `weight_decay`: 0.0 - `adam_beta1`: 0.9 - `adam_beta2`: 0.999 - `adam_epsilon`: 1e-08 - `max_grad_norm`: 1.0 - `num_train_epochs`: 1 - `max_steps`: -1 - `lr_scheduler_type`: linear - `lr_scheduler_kwargs`: {} - `warmup_ratio`: 0.1 - `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`: 12 - `data_seed`: None - `jit_mode_eval`: False - `use_ipex`: False - `bf16`: True - `fp16`: False - `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`: 4 - `dataloader_prefetch_factor`: None - `past_index`: -1 - `disable_tqdm`: False - `remove_unused_columns`: True - `label_names`: None - `load_best_model_at_end`: True - `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} - `deepspeed`: None - `label_smoothing_factor`: 0.0 - `optim`: adamw_torch - `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 - `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 - `dispatch_batches`: None - `split_batches`: 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 - `eval_use_gather_object`: False - `average_tokens_across_devices`: False - `prompts`: None - `batch_sampler`: batch_sampler - `multi_dataset_batch_sampler`: proportional
### Training Logs | Epoch | Step | Training Loss | gooaq-dev_ndcg@10 | NanoMSMARCO_R100_ndcg@10 | NanoNFCorpus_R100_ndcg@10 | NanoNQ_R100_ndcg@10 | NanoBEIR_R100_mean_ndcg@10 | |:----------:|:--------:|:-------------:|:--------------------:|:------------------------:|:-------------------------:|:--------------------:|:--------------------------:| | -1 | -1 | - | 0.1307 (-0.4605) | 0.0867 (-0.4537) | 0.3025 (-0.0226) | 0.0200 (-0.4806) | 0.1364 (-0.3190) | | 0.0001 | 1 | 1.1444 | - | - | - | - | - | | 0.0221 | 200 | 1.182 | - | - | - | - | - | | 0.0443 | 400 | 0.9767 | - | - | - | - | - | | 0.0664 | 600 | 0.5736 | - | - | - | - | - | | 0.0885 | 800 | 0.4752 | - | - | - | - | - | | 0.1106 | 1000 | 0.4281 | 0.7180 (+0.1268) | 0.4989 (-0.0415) | 0.3655 (+0.0405) | 0.5535 (+0.0529) | 0.4726 (+0.0173) | | 0.1328 | 1200 | 0.3803 | - | - | - | - | - | | 0.1549 | 1400 | 0.3646 | - | - | - | - | - | | 0.1770 | 1600 | 0.3535 | - | - | - | - | - | | 0.1992 | 1800 | 0.3498 | - | - | - | - | - | | 0.2213 | 2000 | 0.3237 | 0.7328 (+0.1416) | 0.5173 (-0.0231) | 0.3619 (+0.0368) | 0.6429 (+0.1423) | 0.5074 (+0.0520) | | 0.2434 | 2200 | 0.3199 | - | - | - | - | - | | 0.2655 | 2400 | 0.3196 | - | - | - | - | - | | 0.2877 | 2600 | 0.317 | - | - | - | - | - | | 0.3098 | 2800 | 0.3134 | - | - | - | - | - | | 0.3319 | 3000 | 0.2915 | 0.7501 (+0.1589) | 0.5589 (+0.0184) | 0.3926 (+0.0676) | 0.5667 (+0.0660) | 0.5060 (+0.0507) | | 0.3541 | 3200 | 0.3022 | - | - | - | - | - | | 0.3762 | 3400 | 0.2847 | - | - | - | - | - | | 0.3983 | 3600 | 0.2903 | - | - | - | - | - | | 0.4204 | 3800 | 0.2882 | - | - | - | - | - | | 0.4426 | 4000 | 0.2916 | 0.7516 (+0.1604) | 0.5858 (+0.0454) | 0.3933 (+0.0683) | 0.5949 (+0.0943) | 0.5247 (+0.0693) | | 0.4647 | 4200 | 0.2763 | - | - | - | - | - | | 0.4868 | 4400 | 0.2834 | - | - | - | - | - | | 0.5090 | 4600 | 0.2747 | - | - | - | - | - | | 0.5311 | 4800 | 0.26 | - | - | - | - | - | | 0.5532 | 5000 | 0.2594 | 0.7556 (+0.1643) | 0.5432 (+0.0028) | 0.3748 (+0.0497) | 0.6275 (+0.1268) | 0.5152 (+0.0598) | | 0.5753 | 5200 | 0.273 | - | - | - | - | - | | 0.5975 | 5400 | 0.2616 | - | - | - | - | - | | 0.6196 | 5600 | 0.2573 | - | - | - | - | - | | 0.6417 | 5800 | 0.2426 | - | - | - | - | - | | 0.6639 | 6000 | 0.279 | 0.7605 (+0.1693) | 0.5431 (+0.0026) | 0.3907 (+0.0656) | 0.5926 (+0.0919) | 0.5088 (+0.0534) | | 0.6860 | 6200 | 0.2519 | - | - | - | - | - | | 0.7081 | 6400 | 0.2506 | - | - | - | - | - | | 0.7303 | 6600 | 0.241 | - | - | - | - | - | | 0.7524 | 6800 | 0.2373 | - | - | - | - | - | | 0.7745 | 7000 | 0.2488 | 0.7641 (+0.1728) | 0.5753 (+0.0349) | 0.3897 (+0.0647) | 0.5988 (+0.0981) | 0.5213 (+0.0659) | | 0.7966 | 7200 | 0.2462 | - | - | - | - | - | | 0.8188 | 7400 | 0.2234 | - | - | - | - | - | | 0.8409 | 7600 | 0.235 | - | - | - | - | - | | 0.8630 | 7800 | 0.2209 | - | - | - | - | - | | 0.8852 | 8000 | 0.2267 | 0.7695 (+0.1783) | 0.5509 (+0.0105) | 0.3849 (+0.0598) | 0.5975 (+0.0969) | 0.5111 (+0.0557) | | 0.9073 | 8200 | 0.2322 | - | - | - | - | - | | 0.9294 | 8400 | 0.2273 | - | - | - | - | - | | 0.9515 | 8600 | 0.2111 | - | - | - | - | - | | 0.9737 | 8800 | 0.2371 | - | - | - | - | - | | **0.9958** | **9000** | **0.2328** | **0.7700 (+0.1787)** | **0.5533 (+0.0129)** | **0.3784 (+0.0534)** | **0.5882 (+0.0875)** | **0.5066 (+0.0513)** | | -1 | -1 | - | 0.7700 (+0.1787) | 0.5533 (+0.0129) | 0.3784 (+0.0534) | 0.5882 (+0.0875) | 0.5066 (+0.0513) | * The bold row denotes the saved checkpoint. ### Framework Versions - Python: 3.11.10 - Sentence Transformers: 3.5.0.dev0 - Transformers: 4.49.0 - PyTorch: 2.5.1+cu124 - Accelerate: 1.5.2 - Datasets: 2.21.0 - Tokenizers: 0.21.0 ## Citation ### BibTeX #### Sentence Transformers ```bibtex @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", } ```