Initial commit

.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

NOTICE

@@ -0,0 +1,141 @@
+WaterSheep
+Copyright 2026 Samrat Dutta <samratduttaofficial@gmail.com>
+
+The WaterSheep code and model weights are licensed under the Apache License,
+Version 2.0 (see LICENSE).
+
+-------------------------------------------------------------------------------
+Base model
+-------------------------------------------------------------------------------
+The WaterSheep model is fine-tuned from ModernBERT-base by Answer.AI and
+LightOn (https://huggingface.co/answerdotai/ModernBERT-base), licensed under
+the Apache License, Version 2.0. The encoder weights were modified by
+fine-tuning, and a decision head was added.
+
+-------------------------------------------------------------------------------
+Synthetic training data
+-------------------------------------------------------------------------------
+Synthetic training examples, label descriptions and soft labels were produced
+with Qwen3.5-4B by the Qwen team, Alibaba Cloud
+(https://huggingface.co/Qwen/Qwen3.5-4B), licensed under the Apache License,
+Version 2.0.
+
+-------------------------------------------------------------------------------
+Public training data
+-------------------------------------------------------------------------------
+The model was also trained on the public datasets below. The datasets are not
+included in this repository or in the model files; the code downloads them
+from their sources. Each dataset remains under its own license, as stated by
+its source, and the Apache License of this project does not replace it.
+
+Hugging Face (https://huggingface.co/datasets/<name>)
+  allenai/ai2_arc                                                  CC BY-SA 4.0
+  allenai/openbookqa                                               Apache 2.0
+  allenai/prosocial-dialog                                         CC BY 4.0
+  allenai/quartz                                                   CC BY 4.0
+  allenai/reward-bench                                             ODC-By 1.0
+  allenai/scitail                                                  Apache 2.0
+  allenai/winogrande                                               CC BY
+  Anthropic/hh-rlhf                                                MIT
+  aps/super_glue (COPA)                                            BSD 2-Clause
+  aps/super_glue (WSC)                                             CC BY 4.0
+  benayas/snips                                                    Apache 2.0
+  bitext/Bitext-customer-support-llm-chatbot-training-dataset      CDLA-Sharing 1.0
+  cais/mmlu                                                        MIT
+  chengxuphd/liar2                                                 Apache 2.0
+  clinc/clinc_oos                                                  CC BY 3.0
+  coastalcph/lex_glue                                              CC BY 4.0
+  deepset/prompt-injections                                        Apache 2.0
+  demelin/moral_stories                                            MIT
+  Deysi/spam-detection-dataset                                     Apache 2.0
+  fancyzhx/dbpedia_14                                              CC BY-SA 3.0
+  GBaker/MedQA-USMLE-4-options                                     CC BY 4.0
+  gfissore/arxiv-abstracts-2021                                    CC0 1.0
+  glaiveai/glaive-function-calling-v2                              Apache 2.0
+  gonglinyuan/CoSQA                                                MIT
+  google-research-datasets/go_emotions                             Apache 2.0
+  google-research-datasets/paws                                    Free for any purpose
+  google-research-datasets/poem_sentiment                          CC BY 4.0
+  google/boolq                                                     CC BY-SA 3.0
+  google/civil_comments                                            CC0 1.0
+  gretelai/symptom_to_diagnosis                                    Apache 2.0
+  hendrycks/ethics                                                 MIT
+  HuggingFaceH4/ultrafeedback_binarized                            MIT
+  Intel/orca_dpo_pairs                                             Apache 2.0
+  jackhhao/jailbreak-classification                                Apache 2.0
+  jakartaresearch/semeval-absa                                     CC BY 4.0
+  lmsys/mt_bench_human_judgments                                   CC BY 4.0
+  marksverdhei/clickbait_title_classification                      MIT
+  mikex86/stackoverflow-posts                                      CC BY-SA
+  mmathys/openai-moderation-api-evaluation                         MIT
+  mteb/amazon_counterfactual                                       CC BY 4.0
+  mteb/banking77                                                   MIT
+  mteb/toxic_conversations_50k                                     CC BY 4.0
+  nvidia/Aegis-AI-Content-Safety-Dataset-2.0                       CC BY 4.0
+  nvidia/HelpSteer                                                 CC BY 4.0
+  nvidia/HelpSteer2                                                CC BY 4.0
+  nyu-mll/glue (QNLI)                                              CC BY-SA 4.0
+  nyu-mll/multi_nli                                                OANC / CC BY-SA 3.0 / CC BY 3.0
+  openlifescienceai/medmcqa                                        Apache 2.0
+  owaiskha9654/PubMed_MultiLabel_Text_Classification_Dataset_MeSH  AFL 3.0
+  pminervini/HaluEval                                              Apache 2.0
+  prometheus-eval/Feedback-Collection                              CC BY 4.0
+  prometheus-eval/Preference-Collection                            CC BY 4.0
+  qiaojin/PubMedQA                                                 MIT
+  rajpurkar/squad_v2                                               CC BY-SA 4.0
+  reshabhs/SPML_Chatbot_Prompt_Injection                           MIT
+  SetFit/amazon_massive_intent_en-US                               CC BY 4.0
+  SetFit/amazon_massive_scenario_en-US                             CC BY 4.0
+  SetFit/student-question-categories                               CC0 1.0
+  stanfordnlp/snli                                                 CC BY-SA 4.0
+  tals/vitaminc                                                    CC BY-SA 3.0
+  tasksource/bigbench                                              Apache 2.0
+  tasksource/crowdflower (political media subsets)                 CC0 1.0
+  tasksource/esci                                                  Apache 2.0
+  tasksource/folio                                                 CC BY-SA 4.0
+  tau/commonsense_qa                                               MIT
+  tdavidson/hate_speech_offensive                                  MIT
+  thesofakillers/jigsaw-toxic-comment-classification-challenge     CC BY-SA 3.0
+  TIGER-Lab/MMLU-Pro                                               MIT
+  TimSchopf/medical_abstracts                                      CC BY-SA 3.0
+  truthfulqa/truthful_qa                                           Apache 2.0
+  ucirvine/sms_spam                                                CC BY 4.0
+  zeroshot/twitter-financial-news-sentiment                        MIT
+  zeroshot/twitter-financial-news-topic                            MIT
+
+Kaggle (https://www.kaggle.com/datasets/<name>)
+  andrewmvd/cyberbullying-classification                 CC BY 4.0
+  imoore/60k-stack-overflow-questions-with-quality-rate  MIT / CC BY-SA
+  jp797498e/twitter-entity-sentiment-analysis            CC0 1.0
+  nicapotato/womens-ecommerce-clothing-reviews           CC0 1.0
+  rmisra/clothing-fit-dataset-for-size-recommendation    CC BY 4.0
+  rounakbanik/the-movies-dataset                         CC0 1.0
+  saurabhshahane/ecommerce-text-classification           CC BY 4.0
+  shivamb/real-or-fake-fake-jobposting-prediction        CC0 1.0
+  snap/amazon-fine-food-reviews                          CC0 1.0
+  snehaanbhawal/resume-dataset                           CC0 1.0
+  subhajournal/phishingemails                            LGPL 3.0
+  tboyle10/medicaltranscriptions                         CC0 1.0
+
+Other sources
+  NLU Evaluation Data (HWU64)  CC BY 4.0
+    https://github.com/xliuhw/NLU-Evaluation-Data
+  UCI News Aggregator          CC BY 4.0
+    https://archive.ics.uci.edu/dataset/359/news+aggregator
+  UCI YouTube Spam Collection  CC BY 4.0
+    https://archive.ics.uci.edu/dataset/380/youtube+spam+collection
+
+-------------------------------------------------------------------------------
+Held-out data
+-------------------------------------------------------------------------------
+The pipeline also downloads these datasets but withholds them from training;
+they are used only to test the model on data it has not seen.
+
+  allenai/qasc                                         CC BY 4.0
+    https://huggingface.co/datasets/allenai/qasc
+  PromptCloudHQ/amazon-reviews-unlocked-mobile-phones  CC0 1.0
+    https://www.kaggle.com/datasets/PromptCloudHQ/amazon-reviews-unlocked-mobile-phones
+  rmisra/news-category-dataset                         CC BY 4.0
+    https://www.kaggle.com/datasets/rmisra/news-category-dataset
+  rmisra/news-headlines-dataset-for-sarcasm-detection  CC BY 4.0
+    https://www.kaggle.com/datasets/rmisra/news-headlines-dataset-for-sarcasm-detection

README.md

@@ -0,0 +1,210 @@
+---
+license: apache-2.0
+language:
+- en
+library_name: transformers
+pipeline_tag: zero-shot-classification
+base_model: answerdotai/ModernBERT-base
+tags:
+- decision-model
+- calibration
+- multi-label
+datasets:
+- Anthropic/hh-rlhf
+- Deysi/spam-detection-dataset
+- GBaker/MedQA-USMLE-4-options
+- HuggingFaceH4/ultrafeedback_binarized
+- Intel/orca_dpo_pairs
+- SetFit/amazon_massive_intent_en-US
+- SetFit/amazon_massive_scenario_en-US
+- SetFit/student-question-categories
+- TIGER-Lab/MMLU-Pro
+- TimSchopf/medical_abstracts
+- allenai/ai2_arc
+- allenai/openbookqa
+- allenai/prosocial-dialog
+- allenai/quartz
+- allenai/reward-bench
+- allenai/scitail
+- allenai/winogrande
+- aps/super_glue
+- benayas/snips
+- bitext/Bitext-customer-support-llm-chatbot-training-dataset
+- cais/mmlu
+- chengxuphd/liar2
+- clinc/clinc_oos
+- coastalcph/lex_glue
+- deepset/prompt-injections
+- demelin/moral_stories
+- fancyzhx/dbpedia_14
+- gfissore/arxiv-abstracts-2021
+- glaiveai/glaive-function-calling-v2
+- gonglinyuan/CoSQA
+- google-research-datasets/go_emotions
+- google-research-datasets/paws
+- google-research-datasets/poem_sentiment
+- google/boolq
+- google/civil_comments
+- gretelai/symptom_to_diagnosis
+- hendrycks/ethics
+- jackhhao/jailbreak-classification
+- jakartaresearch/semeval-absa
+- lmsys/mt_bench_human_judgments
+- marksverdhei/clickbait_title_classification
+- mikex86/stackoverflow-posts
+- mmathys/openai-moderation-api-evaluation
+- mteb/amazon_counterfactual
+- mteb/banking77
+- mteb/toxic_conversations_50k
+- nvidia/Aegis-AI-Content-Safety-Dataset-2.0
+- nvidia/HelpSteer
+- nvidia/HelpSteer2
+- nyu-mll/glue
+- nyu-mll/multi_nli
+- openlifescienceai/medmcqa
+- owaiskha9654/PubMed_MultiLabel_Text_Classification_Dataset_MeSH
+- pminervini/HaluEval
+- prometheus-eval/Feedback-Collection
+- prometheus-eval/Preference-Collection
+- qiaojin/PubMedQA
+- rajpurkar/squad_v2
+- reshabhs/SPML_Chatbot_Prompt_Injection
+- stanfordnlp/snli
+- tals/vitaminc
+- tasksource/bigbench
+- tasksource/crowdflower
+- tasksource/esci
+- tasksource/folio
+- tau/commonsense_qa
+- tdavidson/hate_speech_offensive
+- thesofakillers/jigsaw-toxic-comment-classification-challenge
+- truthfulqa/truthful_qa
+- ucirvine/sms_spam
+- zeroshot/twitter-financial-news-sentiment
+- zeroshot/twitter-financial-news-topic
+---
+
+![WaterSheep](https://samratduttaofficial.github.io/WaterSheep/banner.svg)
+
+# ![](https://samratduttaofficial.github.io/WaterSheep/logo.svg) WaterSheep
+
+[Website](https://samratduttaofficial.github.io/WaterSheep/) · [Demo](https://huggingface.co/spaces/samratduttaofficial/WaterSheep) · [Code](https://github.com/SamratDuttaOfficial/WaterSheep)
+
+WaterSheep answers yes/no, single-choice, rating and multi-label questions about any text, with a
+probability for every option. Version 0.1.0 (`watersheep-20260928-125452`).
+
+## Usage
+
+```bash
+pip install transformers torch
+```
+
+```python
+from transformers import pipeline
+
+ws = pipeline(model="samratduttaofficial/WaterSheep", trust_remote_code=True)
+ws("I was charged twice.", question="Which team should handle this?", options=["billing", "shipping", "support"])
+```
+
+| Type | Options | Answer |
+|---|---|---|
+| `noul` | none (yes/no) | probability of yes |
+| `choice` | any labels | the best option |
+| `score` | a digit scale, e.g. `1` to `5` | the expected level |
+| `multi` | any labels, with `type="multi"` | every option above the threshold |
+
+Every answer includes a probability for each option.
+
+## Download
+
+```bash
+hf download samratduttaofficial/WaterSheep --local-dir WaterSheep
+```
+
+Or with Git (requires Git LFS):
+
+```bash
+git clone https://huggingface.co/samratduttaofficial/WaterSheep
+```
+
+Then load it from the folder, offline:
+
+```python
+ws = pipeline(model="WaterSheep", trust_remote_code=True)
+```
+
+## API
+
+Deploy it as an [Inference Endpoint](https://endpoints.huggingface.co), then:
+
+```bash
+curl https://YOUR-ENDPOINT -H "Authorization: Bearer $HF_TOKEN" -H "Content-Type: application/json" -d '{"inputs": "I was charged twice.", "parameters": {"question": "Which team should handle this?", "options": ["billing", "shipping", "support"]}}'
+```
+
+## JavaScript
+
+No install; runs in the browser:
+
+```html
+<script type="module">
+  import { decide } from "https://samratduttaofficial.github.io/WaterSheep/watersheep.js";
+  console.log(await decide("I was charged twice.", "Which team should handle this?", ["billing", "shipping", "support"]));
+</script>
+```
+
+With a downloaded copy on your web server, call `load({ base: "WaterSheep/" })` first.
+
+Other languages: run `onnx/model_quantized.onnx` with ONNX Runtime; `watersheep.js` shows the input format.
+
+## Evaluation
+
+| Evaluation | Accuracy | ECE |
+|---|---|---|
+| In-distribution test split | 77.8% | 0.026 |
+| Held-out datasets, not seen in training | 61.2% | 0.043 |
+
+ECE is the expected calibration error (lower is better).
+
+### Benchmarks
+
+| Benchmark | Suite | Questions | Accuracy | ECE | In training data |
+|---|---|---|---|---|---|
+| [goemotions](https://huggingface.co/datasets/google-research-datasets/go_emotions) | sentiment | 2000 | 22.4% | 0.023 | other split |
+| [hatecheck](https://huggingface.co/datasets/Paul/hatecheck) | safety | 2000 | 75.1% | 0.139 | no |
+| [legal_abercrombie](https://huggingface.co/datasets/nguha/legalbench) | legal | 95 | 21.1% | 0.316 | no |
+| [legal_contract_nli_confidentiality_of_agreement](https://huggingface.co/datasets/nguha/legalbench) | legal | 82 | 69.5% | 0.177 | no |
+| [legal_corporate_lobbying](https://huggingface.co/datasets/nguha/legalbench) | legal | 490 | 68.4% | 0.216 | no |
+| [legal_cuad_audit_rights](https://huggingface.co/datasets/nguha/legalbench) | legal | 1216 | 86.3% | 0.041 | no |
+| [legal_definition_classification](https://huggingface.co/datasets/nguha/legalbench) | legal | 1337 | 56.9% | 0.279 | no |
+| [legal_function_of_decision_section](https://huggingface.co/datasets/nguha/legalbench) | legal | 367 | 24.3% | 0.245 | no |
+| [legal_hearsay](https://huggingface.co/datasets/nguha/legalbench) | legal | 94 | 56.4% | 0.307 | no |
+| [legal_overruling](https://huggingface.co/datasets/nguha/legalbench) | legal | 2000 | 62.5% | 0.151 | no |
+| [legal_personal_jurisdiction](https://huggingface.co/datasets/nguha/legalbench) | legal | 50 | 50.0% | 0.160 | no |
+| [legal_privacy_policy_qa](https://huggingface.co/datasets/nguha/legalbench) | legal | 2000 | 58.9% | 0.274 | no |
+| [legal_proa](https://huggingface.co/datasets/nguha/legalbench) | legal | 95 | 51.6% | 0.379 | no |
+| [legal_ucc_v_common_law](https://huggingface.co/datasets/nguha/legalbench) | legal | 94 | 62.8% | 0.171 | no |
+| [prompt_injection](https://huggingface.co/datasets/deepset/prompt-injections) | safety | 116 | 91.4% | 0.079 | other split |
+| [xstest](https://huggingface.co/datasets/Paul/XSTest) | safety | 450 | 73.6% | 0.140 | no |
+
+## Limitations
+
+- English only.
+- Long inputs are truncated.
+- Rating-scale answers are less accurate than the other types.
+- Probabilities are calibrated on data like the training data; validate them on your own.
+- Not for high-stakes decisions (medical, legal, financial, hiring) on its own.
+
+## License
+
+Apache 2.0 (`LICENSE`). Trained on openly licensed data; credits in `NOTICE`.
+
+## Citation
+
+```bibtex
+@misc{watersheep,
+  author = {Samrat Dutta},
+  title  = {WaterSheep: calibrated decisions for any text},
+  year   = {2026},
+  url    = {https://huggingface.co/samratduttaofficial/WaterSheep}
+}
+```

config.json

@@ -0,0 +1,109 @@
+{
+  "architectures": [
+    "WaterSheepModel"
+  ],
+  "model_type": "watersheep",
+  "auto_map": {
+    "AutoConfig": "modeling_watersheep.WaterSheepConfig",
+    "AutoModel": "modeling_watersheep.WaterSheepModel"
+  },
+  "custom_pipelines": {
+    "zero-shot-classification": {
+      "impl": "pipeline_watersheep.WaterSheepPipeline",
+      "pt": [
+        "AutoModel"
+      ]
+    }
+  },
+  "encoder_config": {
+    "architectures": [
+      "ModernBertForMaskedLM"
+    ],
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "bos_token_id": 50281,
+    "classifier_activation": "gelu",
+    "classifier_bias": false,
+    "classifier_dropout": 0.0,
+    "classifier_pooling": "mean",
+    "cls_token_id": 50281,
+    "decoder_bias": true,
+    "deterministic_flash_attn": false,
+    "dtype": "float32",
+    "embedding_dropout": 0.0,
+    "eos_token_id": 50282,
+    "global_attn_every_n_layers": 3,
+    "gradient_checkpointing": false,
+    "hidden_activation": "gelu",
+    "hidden_size": 768,
+    "initializer_cutoff_factor": 2.0,
+    "initializer_range": 0.02,
+    "intermediate_size": 1152,
+    "layer_norm_eps": 1e-05,
+    "layer_types": [
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention"
+    ],
+    "local_attention": 128,
+    "max_position_embeddings": 8192,
+    "mlp_bias": false,
+    "mlp_dropout": 0.0,
+    "model_type": "modernbert",
+    "norm_bias": false,
+    "norm_eps": 1e-05,
+    "num_attention_heads": 12,
+    "num_hidden_layers": 22,
+    "pad_token_id": 50283,
+    "position_embedding_type": "absolute",
+    "rope_parameters": {
+      "full_attention": {
+        "rope_theta": 160000.0,
+        "rope_type": "default"
+      },
+      "sliding_attention": {
+        "rope_theta": 10000.0,
+        "rope_type": "default"
+      }
+    },
+    "sep_token_id": 50282,
+    "sparse_pred_ignore_index": -100,
+    "sparse_prediction": false,
+    "tie_word_embeddings": true,
+    "transformers_version": "5.16.1",
+    "vocab_size": 50368
+  },
+  "head_layers": 1,
+  "max_len": 512,
+  "max_question_tokens": 96,
+  "max_option_tokens": 32,
+  "max_options": 10,
+  "temperatures": {
+    "binary": 1.0594,
+    "choice": 1.0631,
+    "score": 1.0022,
+    "multi": 0.8276
+  },
+  "multi_threshold": 0.5,
+  "dtype": "float32",
+  "transformers_version": "5.16.1"
+}

encoder/config.json

@@ -0,0 +1,77 @@
+{
+  "architectures": [
+    "ModernBertForMaskedLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 50281,
+  "classifier_activation": "gelu",
+  "classifier_bias": false,
+  "classifier_dropout": 0.0,
+  "classifier_pooling": "mean",
+  "cls_token_id": 50281,
+  "decoder_bias": true,
+  "deterministic_flash_attn": false,
+  "dtype": "float32",
+  "embedding_dropout": 0.0,
+  "eos_token_id": 50282,
+  "global_attn_every_n_layers": 3,
+  "gradient_checkpointing": false,
+  "hidden_activation": "gelu",
+  "hidden_size": 768,
+  "initializer_cutoff_factor": 2.0,
+  "initializer_range": 0.02,
+  "intermediate_size": 1152,
+  "layer_norm_eps": 1e-05,
+  "layer_types": [
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention"
+  ],
+  "local_attention": 128,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "mlp_dropout": 0.0,
+  "model_type": "modernbert",
+  "norm_bias": false,
+  "norm_eps": 1e-05,
+  "num_attention_heads": 12,
+  "num_hidden_layers": 22,
+  "pad_token_id": 50283,
+  "position_embedding_type": "absolute",
+  "rope_parameters": {
+    "full_attention": {
+      "rope_theta": 160000.0,
+      "rope_type": "default"
+    },
+    "sliding_attention": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    }
+  },
+  "sep_token_id": 50282,
+  "sparse_pred_ignore_index": -100,
+  "sparse_prediction": false,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.16.1",
+  "vocab_size": 50368
+}

handler.py

@@ -0,0 +1,9 @@
+from transformers import pipeline
+
+
+class EndpointHandler:
+    def __init__(self, path=""):
+        self.pipe = pipeline(model=path, trust_remote_code=True)
+
+    def __call__(self, data):
+        return self.pipe(data["inputs"], **(data.get("parameters") or {}))

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e55d28e1a26fada5ab701fce713ad56cc1d3d849df2d8be79c1a11a08da0981a
+size 617352620

modeling_watersheep.py

@@ -0,0 +1,57 @@
+from dataclasses import dataclass
+
+import torch
+from torch import nn
+from transformers import AutoConfig, AutoModel, PreTrainedConfig, PreTrainedModel
+from transformers.utils import ModelOutput
+
+
+class WaterSheepConfig(PreTrainedConfig):
+    model_type = "watersheep"
+
+    def __init__(self, encoder_config=None, head_layers=1, max_len=512, max_question_tokens=96,
+                 max_option_tokens=32, max_options=10, temperatures=None, multi_threshold=0.5, **kwargs):
+        self.encoder_config = encoder_config or {}
+        self.head_layers = head_layers
+        self.max_len = max_len
+        self.max_question_tokens = max_question_tokens
+        self.max_option_tokens = max_option_tokens
+        self.max_options = max_options
+        self.temperatures = temperatures or {}
+        self.multi_threshold = multi_threshold
+        super().__init__(**kwargs)
+
+
+@dataclass
+class WaterSheepOutput(ModelOutput):
+    logits: torch.FloatTensor = None
+
+
+class WaterSheepModel(PreTrainedModel):
+    config_class = WaterSheepConfig
+    base_model_prefix = "watersheep"
+
+    def __init__(self, config):
+        super().__init__(config)
+        enc = AutoConfig.for_model(**config.encoder_config)
+        if hasattr(enc, "reference_compile"):
+            enc.reference_compile = False
+        self.encoder = AutoModel.from_config(enc)
+        h = enc.hidden_size
+        self.proj = nn.Sequential(nn.Dropout(0.0), nn.Linear(h, h), nn.GELU(), nn.LayerNorm(h))
+        self.mix = None
+        if config.head_layers > 0:
+            layer = nn.TransformerEncoderLayer(h, nhead=max(1, h // 64), dim_feedforward=2 * h, dropout=0.0,
+                                               activation="gelu", batch_first=True, norm_first=True)
+            self.mix = nn.TransformerEncoder(layer, config.head_layers, enable_nested_tensor=False)
+        self.out = nn.Linear(h, 1)
+        self.post_init()
+
+    def forward(self, input_ids, attention_mask, option_positions, option_mask, **kwargs):
+        hs = self.encoder(input_ids=input_ids, attention_mask=attention_mask).last_hidden_state
+        idx = option_positions.clamp(min=0).unsqueeze(-1).expand(-1, -1, hs.size(-1))
+        x = self.proj(torch.gather(hs, 1, idx))
+        if self.mix is not None:
+            x = self.mix(x, src_key_padding_mask=~option_mask)
+        logits = self.out(x).squeeze(-1).float()
+        return WaterSheepOutput(logits=logits.masked_fill(~option_mask, -1e4))

onnx/model_quantized.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:160ae91868f919ae0ddc6d1bd509b17d93194102aa33a466095469fbf04dc25b
+size 158196995

pipeline_watersheep.py

@@ -0,0 +1,179 @@
+import numpy as np
+import torch
+from transformers import AutoTokenizer, Pipeline
+
+TYPES = {"noul": "binary", "binary": "binary", "yes/no": "binary", "boolean": "binary",
+         "choice": "choice", "score": "score", "multi": "multi", "multi_choice": "multi",
+         "multilabel": "multi", "multi-label": "multi"}
+TAGS = {"binary": "[yes/no]", "choice": "[choose]", "score": "[rate]", "multi": "[select all]"}
+DEFAULT_QUESTION = {"choice": "Which label fits the text?", "multi": "Which labels apply to the text?"}
+YESNO = ["yes", "no"]
+
+
+def _softmax(z, t):
+    z = np.asarray(z, np.float64) / max(1e-6, t)
+    e = np.exp(z - z.max())
+    return e / e.sum()
+
+
+def _sigmoid(z, t):
+    z = np.clip(np.asarray(z, np.float64) / max(1e-6, t), -60, 60)
+    return 1.0 / (1.0 + np.exp(-z))
+
+
+def _digits(opts):
+    return all(len(o) == 1 and o.isdigit() for o in opts)
+
+
+def _infer_type(opts):
+    low = [o.strip().lower() for o in opts]
+    if low == YESNO:
+        return "binary"
+    if _digits(low):
+        v = [int(o) for o in low]
+        if v == list(range(v[0], v[0] + len(v))):
+            return "score"
+    return "choice"
+
+
+class WaterSheepPipeline(Pipeline):
+    _load_tokenizer = False
+    _load_processor = False
+    _load_image_processor = False
+    _load_video_processor = False
+    _load_feature_extractor = False
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        if self.tokenizer is None:
+            self.tokenizer = AutoTokenizer.from_pretrained(
+                self.model.name_or_path, subfolder="tokenizer",
+                revision=getattr(self.model.config, "_commit_hash", None))
+
+    def _sanitize_parameters(self, question=None, options=None, candidate_labels=None, type=None,
+                             multi_label=None, threshold=None, hypothesis_template=None):
+        pre = {}
+        if question is not None:
+            pre["question"] = question
+        opts = options if options is not None else candidate_labels
+        if isinstance(opts, str):
+            opts = [o.strip() for o in opts.split(",") if o.strip()]
+        if opts is not None:
+            pre["options"] = list(opts)
+        if type is not None or multi_label:
+            pre["type"] = type or "multi"
+        return pre, {}, {} if threshold is None else {"threshold": threshold}
+
+    def preprocess(self, text, question=None, options=None, type=None):
+        opts = [str(o) for o in (options or YESNO)]
+        t = TYPES.get(str(type or "").lower()) or _infer_type(opts)
+        if t == "binary":
+            opts = list(YESNO)
+        question = question or DEFAULT_QUESTION.get(t)
+        if not question:
+            raise ValueError("a question is required")
+        return {"text": "" if text is None else str(text), "question": str(question), "type": t, "options": opts}
+
+    def _forward(self, x):
+        return {**x, "probs": self._probs(x["type"], x["text"], x["question"], x["options"])}
+
+    def postprocess(self, x, threshold=None):
+        t, opts, p = x["type"], x["options"], x["probs"]
+        res = {"type": t, "probs": {o: float(v) for o, v in zip(opts, p)}}
+        if t == "multi":
+            thr = self.model.config.multi_threshold if threshold is None else float(threshold)
+            res.update(answer=[o for o, v in zip(opts, p) if v >= thr],
+                       confidence=float(np.mean([max(v, 1 - v) for v in p])))
+        else:
+            i = int(np.argmax(p))
+            res.update(answer=opts[i], index=i, confidence=float(p[i]))
+        if t == "binary":
+            res["p_yes"] = float(p[0])
+        elif t == "score":
+            res["expected"] = float(sum(k * v for k, v in enumerate(p))) + (int(opts[0]) if _digits(opts) else 0)
+        return res
+
+    def _logits(self, t, text, question, groups):
+        cfg, tok = self.model.config, self.tokenizer
+        cls = tok.cls_token_id if tok.cls_token_id is not None else tok.bos_token_id
+        sep = tok.sep_token_id if tok.sep_token_id is not None else tok.eos_token_id
+        mask, pad = tok.mask_token_id, tok.pad_token_id if tok.pad_token_id is not None else 0
+        enc = lambda xs: tok(xs, add_special_tokens=False)["input_ids"]
+        q = enc([TAGS[t] + " " + question])[0][:cfg.max_question_tokens]
+        s = enc([text])[0]
+        flat = enc([o for g in groups for o in g])
+        items, k = [], 0
+        for g in groups:
+            ids, pos = [cls] + q + [sep], []
+            for o in flat[k:k + len(g)]:
+                pos.append(len(ids))
+                ids += [mask] + o[:cfg.max_option_tokens]
+            k += len(g)
+            ids.append(sep)
+            room = cfg.max_len - len(ids) - 1
+            if room > 0 and s:
+                ids += s[:room]
+            ids.append(sep)
+            if len(ids) > cfg.max_len:
+                ids = ids[:cfg.max_len - 1] + [sep]
+                pos = [p for p in pos if p < cfg.max_len - 1]
+            items.append((ids, pos))
+        out, dev = [], self.device
+        bf16 = dev.type == "cuda" and torch.cuda.is_bf16_supported()
+        for i in range(0, len(items), 64):
+            part = items[i:i + 64]
+            n, m = max(len(a) for a, _ in part), max(len(p) for _, p in part)
+            x = torch.full((len(part), n), pad, dtype=torch.long)
+            am = torch.zeros((len(part), n), dtype=torch.long)
+            op = torch.zeros((len(part), m), dtype=torch.long)
+            om = torch.zeros((len(part), m), dtype=torch.bool)
+            for b, (ids, pos) in enumerate(part):
+                x[b, :len(ids)] = torch.tensor(ids)
+                am[b, :len(ids)] = 1
+                op[b, :len(pos)] = torch.tensor(pos)
+                om[b, :len(pos)] = True
+            with torch.autocast(dev.type, dtype=torch.bfloat16, enabled=bf16):
+                z = self.model(x.to(dev), am.to(dev), op.to(dev), om.to(dev)).logits.float().cpu().numpy()
+            for (ids, pos), g, zz in zip(part, groups[i:i + 64], z):
+                out.append(zz[:len(pos)] if len(pos) == len(g) else None)
+        return out
+
+    def _probs(self, t, text, question, opts, size=0):
+        cfg = self.model.config
+        temp = (cfg.temperatures or {}).get(t, 1.0)
+        n = len(opts)
+        size = min(size or max(2, min(10, int(cfg.max_options or 10))), n)
+        while True:
+            if n <= size:
+                (z,) = self._logits(t, text, question, [opts])
+                if z is not None:
+                    return _sigmoid(z, temp) if t == "multi" else _softmax(z, temp)
+            else:
+                groups = [list(range(k, min(n, k + size))) for k in range(0, n, size)]
+                zs = self._logits(t, text, question, [[opts[i] for i in g] for g in groups])
+                if all(z is not None for z in zs):
+                    break
+            if size <= 2:
+                raise ValueError("the options do not fit in the model's input - shorten them")
+            size = max(2, min(size - 1, 8) if size > 8 else size // 2)
+        if t == "multi":
+            p = np.zeros(n)
+            for g, z in zip(groups, zs):
+                p[g] = _sigmoid(z, temp)
+            return p
+        keep = max(1, size // len(groups))
+        local, finalists = {}, []
+        for g, z in zip(groups, zs):
+            pg = _softmax(z, temp)
+            local.update({i: pg[k] for k, i in enumerate(g)})
+            finalists += [g[k] for k in np.argsort(-pg)[:keep]]
+        pf = self._probs(t, text, question, [opts[i] for i in finalists], size)
+        p = np.zeros(n)
+        for g in groups:
+            top = max((i for i in g if i in finalists), key=lambda i: local[i])
+            scale = pf[finalists.index(top)] / max(1e-12, local[top])
+            for i in g:
+                p[i] = local[i] * scale
+        for k, i in enumerate(finalists):
+            p[i] = pf[k]
+        return p / p.sum()

requirements.txt

@@ -0,0 +1 @@
+transformers>=5

tokenizer/tokenizer_config.json

@@ -0,0 +1,17 @@
+{
+  "backend": "tokenizers",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "[MASK]",
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 8192,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": "[UNK]"
+}

watersheep.json

@@ -0,0 +1,30 @@
+{
+  "format": 2,
+  "run_id": "run_20260928-070402",
+  "encoder": "answerdotai/ModernBERT-base",
+  "tokenizer": "hf",
+  "max_len": 512,
+  "max_question_tokens": 96,
+  "max_option_tokens": 32,
+  "max_options": 10,
+  "head_layers": 1,
+  "temperatures": {
+    "binary": 1.0594,
+    "choice": 1.0631,
+    "score": 1.0022,
+    "multi": 0.8276
+  },
+  "metrics": {
+    "test_calibrated": {
+      "acc": 0.7778176740474766,
+      "ece": 0.025550670609045353
+    },
+    "zeroshot_calibrated": {
+      "acc": 0.6121272507294855,
+      "ece": 0.04295260971909359
+    }
+  },
+  "multi_threshold": 0.5,
+  "created": "20260928-125453",
+  "name": "watersheep-20260928-125452"
+}