[docs] Use modality-neutral terms (input, document) in loss docs & docstrings

[tomaarsen]

Hello!

Pull Request overview

• Replace text-centric wording in the loss overview docs and loss docstrings: sentence/sentences -> input/inputs, passage/passages -> document/documents
• Rename the Texts column to Inputs in the loss input-format tables (both the loss_overview.md pages and the docstring Inputs: grid tables)

Details

Now that SentenceTransformer, CrossEncoder, and SparseEncoder models are now multimodal, a lot of the documentation framing reads as too text-only. Tables that describe the accepted training data formats call the input column Texts and spell out formats like (sentence_A, sentence_B) pairs or single sentences, and a fair amount of prose talks about "sentences" and "passages" where it really just means "inputs" and "documents".

This PR sweeps that terminology to be modality-neutral:

• docs/sentence_transformer/loss_overview.md, docs/cross_encoder/loss_overview.md, docs/sparse_encoder/loss_overview.md: Texts -> Inputs headers, (sentence_A, sentence_B) pairs -> (input_A, input_B) pairs, single sentences -> single inputs, (damaged_sentence, original_sentence) pairs -> (damaged_input, original_input) pairs, (query, passage_one, passage_two) -> (query, document_one, document_two), teacher/model sentence embeddings -> teacher/model embeddings, plus the matching prose.

• All loss docstrings under sentence_transformers/{sentence_transformer,cross_encoder,sparse_encoder}/losses/: the Inputs: grid tables and the Requirements:/description prose, with the RST grid borders re-rendered to fit.

• sentence_transformers/util/hard_negatives.py: the mine_hard_negatives docstring, so its (anchor, passage, label) / (anchor, [passages], …) output-format descriptions line up with the rest.

• Tom Aarsen

[Copilot]

Pull request overview

Updates documentation across SentenceTransformer, CrossEncoder, and SparseEncoder loss docs/docstrings to use modality-neutral terminology now that models can be multimodal.

Changes:

• Replaces text-centric terms (e.g., “sentence(s)”, “passage(s)”, “Texts”) with modality-neutral terms (“input(s)”, “document(s)”, “Inputs”) in loss overviews and loss docstrings.
• Updates example input-format tables (Markdown + RST grid tables) to reflect the new terminology.
• Aligns mine_hard_negatives docstring output-format terminology with “document”/“inputs”.

Reviewed changes

Copilot reviewed 54 out of 54 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
sentence_transformers/util/hard_negatives.py	Updates hard-negative mining docstring to use “inputs”/“documents” terminology.
sentence_transformers/sparse_encoder/losses/sparse_triplet.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sparse_encoder/losses/sparse_multiple_negatives_ranking.py	Makes docstring wording modality-neutral (“embeddings”, “Inputs”).
sentence_transformers/sparse_encoder/losses/sparse_mse.py	Updates distillation wording/table from “sentence” → “input” and “sentence embeddings” → “embeddings”.
sentence_transformers/sparse_encoder/losses/sparse_margin_mse.py	Updates “passage” → “document” and “Texts” → “Inputs” in docstring.
sentence_transformers/sparse_encoder/losses/sparse_distill_kl_div.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sparse_encoder/losses/sparse_cosine_similarity.py	Updates docstring examples to use input_A/input_B + “Inputs” table header.
sentence_transformers/sparse_encoder/losses/sparse_cosent.py	Updates docstring to “Input pairs” + “Inputs” table header.
sentence_transformers/sparse_encoder/losses/sparse_angle.py	Updates docstring to “Input pairs” + “Inputs” table header.
sentence_transformers/sparse_encoder/losses/csr.py	Updates docstrings to refer to “embeddings”/“inputs” instead of “sentence embeddings”/“sentences”.
sentence_transformers/sparse_encoder/losses/cached_splade.py	Updates helper docstrings from “sentences” → “inputs”.
sentence_transformers/sentence_transformer/losses/triplet.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/softmax.py	Updates requirements + table wording from sentence pairs → input pairs.
sentence_transformers/sentence_transformer/losses/online_contrastive.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/multiple_negatives_symmetric_ranking.py	Updates docstring to “embeddings” + “Inputs” table header.
sentence_transformers/sentence_transformer/losses/multiple_negatives_ranking.py	Updates docstring to “embeddings” + “Inputs” table header.
sentence_transformers/sentence_transformer/losses/mse.py	Updates distillation wording/table from “sentence embedding(s)” → “embedding(s)” and “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/mega_batch_margin.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/matryoshka.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/matryoshka_2d.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/margin_mse.py	Updates “passage” → “document”, “Texts” → “Inputs”, and “sentences” → “inputs” in docstring.
sentence_transformers/sentence_transformer/losses/global_orthogonal_regularization.py	Updates docstring tables and examples (“passages” → “documents”, “Texts” → “Inputs”).
sentence_transformers/sentence_transformer/losses/gist_embed.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/embed_distill.py	Updates distillation table wording from “sentence” → “input”.
sentence_transformers/sentence_transformer/losses/distill_kl_div.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/denoising_auto_encoder.py	Updates TSDAE loss docstring wording from sentences → inputs + updates input table examples.
sentence_transformers/sentence_transformer/losses/cosine_similarity.py	Updates docstring examples to use input_A/input_B + “Inputs” table header.
sentence_transformers/sentence_transformer/losses/cosent.py	Updates docstring to “Input pairs” + “Inputs” table header.
sentence_transformers/sentence_transformer/losses/contrastive.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/contrastive_tension.py	Updates docstrings/examples/tables from sentences → inputs and “sentence embeddings” → “embeddings”.
sentence_transformers/sentence_transformer/losses/cached_multiple_negatives_symmetric_ranking.py	Updates docstring to “embeddings” + “Inputs” table header.
sentence_transformers/sentence_transformer/losses/cached_multiple_negatives_ranking.py	Updates docstring to “embeddings” + “Inputs” table header and helper docstring (“sentences” → “inputs”).
sentence_transformers/sentence_transformer/losses/cached_gist_embed.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/sentence_transformer/losses/batch_semi_hard_triplet.py	Updates docstring wording/tables from sentences → inputs.
sentence_transformers/sentence_transformer/losses/batch_hard_triplet.py	Updates docstring wording/tables from sentences → inputs.
sentence_transformers/sentence_transformer/losses/batch_hard_soft_margin_triplet.py	Updates docstring wording/tables from sentences → inputs.
sentence_transformers/sentence_transformer/losses/batch_all_triplet.py	Updates docstring wording/tables from sentences → inputs.
sentence_transformers/sentence_transformer/losses/angle.py	Updates docstring to “Input pairs” + “Inputs” table header.
sentence_transformers/sentence_transformer/losses/adaptive_layer.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/rank_net.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/plist_mle.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/multiple_negatives_ranking.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/mse.py	Updates wording from query-passage → query-document and “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/margin_mse.py	Updates wording from passages → documents and “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/list_net.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/list_mle.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/lambda_loss.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/cross_entropy.py	Updates wording from sentence pairs → input pairs and “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/cached_multiple_negatives_ranking.py	Renames docstring table column “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/binary_cross_entropy.py	Updates table examples from sentence_A/sentence_B → input_A/input_B and “Texts” → “Inputs”.
sentence_transformers/cross_encoder/losses/adr_mse.py	Renames docstring table column “Texts” → “Inputs”.
docs/sparse_encoder/loss_overview.md	Updates sparse loss overview tables/prose to modality-neutral terminology.
docs/sentence_transformer/loss_overview.md	Updates sentence-transformer loss overview tables/prose to modality-neutral terminology.
docs/cross_encoder/loss_overview.md	Updates cross-encoder loss overview tables/prose (incl. mine_hard_negatives references) to modality-neutral terminology.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 54 out of 54 changed files in this pull request and generated 1 comment.

[Copilot]

Pull request overview

Copilot reviewed 54 out of 54 changed files in this pull request and generated 5 comments.