Duplicate from Adam-12-0/omnitom-benchmark

.gitattributes

@@ -0,0 +1,60 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro 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
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds 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
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Adam Bawatneh, Sagar Sapkota, Amrit Singh Bedi, Santu Karmaker, and Mubarak Shah
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,216 @@
+---
+pretty_name: OmniToM
+language:
+- en
+license: mit
+task_categories:
+- text-generation
+- text-classification
+task_ids:
+- text2text-generation
+- multi-class-classification
+size_categories:
+- n<1K
+tags:
+- theory-of-mind
+- benchmark
+- social-reasoning
+- belief-modeling
+- llm-evaluation
+- cognitive-reasoning
+- mental-state-representation
+---
+
+# Dataset Card for OmniToM
+
+## Dataset Details
+
+### Dataset Description
+
+OmniToM is a benchmark for evaluating Theory of Mind in language models through explicit belief-structure modeling. Instead of scoring only endpoint answers to social-reasoning questions, OmniToM exposes the intermediate belief structure that a model must build in order to reason coherently about what different actors know, believe, infer, intend, or misunderstand.
+
+Each example is a short English story paired with:
+
+- a set of actor-centered belief propositions
+- a reserved `world` actor for narrator/world facts
+- a seven-dimensional schema label vector for every belief
+
+The benchmark supports two linked tasks:
+
+1. **Belief Extraction**
+   - Given a story, extract the relevant belief structure as `(Actor, Belief, Order)` tuples.
+2. **Belief Labeling**
+   - Given the story and belief tuples, label each belief along seven closed-set schema dimensions.
+
+- **Language(s):** English
+- **License:** MIT
+
+## Uses
+
+### Direct Use
+
+OmniToM is designed for benchmark evaluation and diagnostic analysis. Suitable uses include:
+
+- zero-shot belief extraction
+- zero-shot belief labeling
+- semantic-judge evaluation of extracted belief tables
+- analysis of multi-actor and higher-order Theory-of-Mind reasoning
+- process-sensitive evaluation beyond endpoint question answering
+
+### Out-of-Scope Use
+
+OmniToM should not be treated as:
+
+- a direct measure of real-world social intelligence
+- a measure of embodied, interactive, or multimodal social reasoning
+- a safety certification benchmark for deployed systems
+- a clinical, educational, or psychological assessment tool
+- a complete coverage benchmark for all possible Theory-of-Mind phenomena
+
+## Dataset Structure
+
+### Data Instances
+
+Each line in the release file is one JSON object:
+
+```json
+{
+  "story_id": 1,
+  "story_category": "Ambiguous Story Task",
+  "story": "Story text...",
+  "beliefs": [
+    {
+      "actor": "world",
+      "belief": "A minimal propositional statement.",
+      "labels": {
+        "order": "0",
+        "truth_status": "True",
+        "knowledge_access": "Public",
+        "representation": "Explicit",
+        "content_type": "Action/Event",
+        "mental_source": "Narration",
+        "context": "Neutral"
+      }
+    }
+  ]
+}
+```
+
+### Data Fields
+
+- `story_id`
+  - Unique integer identifier for the story.
+- `story_category`
+  - One of seven retained benchmark categories.
+- `story`
+  - Raw story text used for both extraction and labeling tasks.
+- `beliefs`
+  - List of annotated belief propositions.
+- `beliefs[].actor`
+  - Belief holder. The reserved actor `world` denotes narrator/world facts.
+- `beliefs[].belief`
+  - Minimal propositional belief statement.
+- `beliefs[].labels.order`
+  - Recursive depth in `{0,1,2,3}`.
+- `beliefs[].labels.truth_status`
+  - `True`, `False`, or `Unknown`.
+- `beliefs[].labels.knowledge_access`
+  - `Private`, `Shared`, or `Public`.
+- `beliefs[].labels.representation`
+  - `Explicit` or `Implicit`.
+- `beliefs[].labels.content_type`
+  - One of: `Location`, `Contents/Physical State`, `Identity/Relation`, `Epistemic`, `Desire/Intention`, `Emotion`, `Trait/Value`, `Action/Event`.
+- `beliefs[].labels.mental_source`
+  - One of: `Narration`, `Perception`, `Memory`, `Testimony`, `Inference`, `Imagination`, `Unknown`.
+- `beliefs[].labels.context`
+  - `Neutral`, `Temporal`, `Deceptive`, or `Counterfactual`.
+
+### Data Splits
+
+This release contains one benchmark split:
+
+- `train` / benchmark split: `895` stories
+
+The split is named `train` in the Hugging Face dataset viewer for compatibility with the default dataset loading interface. It should be interpreted as the benchmark split, not as a recommended training set.
+
+## Dataset Creation
+
+### Curation Rationale
+
+OmniToM was created to address a limitation in prior Theory-of-Mind benchmarks for language models: most evaluate endpoint question answering rather than whether a model constructs a coherent belief representation while reading the story.
+
+OmniToM instead evaluates explicit belief-structure modeling. The benchmark is grounded in short stories from ToMBench and organizes reasoning around an ATOMS-grounded belief-level schema for fine-grained analysis of mental-state representations.
+
+### Source Data
+
+The benchmark sources its stories from ToMBench. From the original source corpus, OmniToM retains seven story categories whose stories provide sufficiently self-contained mental-state evidence for belief extraction from text alone:
+
+- Ambiguous Story Task
+- False Belief Task
+- Faux-pas Recognition Test
+- Hinting Task Test
+- Persuasion Story Task
+- Scalar Implicature Test
+- Strange Story Task
+
+### Annotation Process
+
+The benchmark was built with a human-validated, LLM-assisted annotation pipeline:
+
+- `1,383` source stories in the original corpus
+- `916` stories retained after source filtering
+- `895` stories released in the final benchmark
+- `22,343` labeled belief propositions in the released benchmark
+- `156,401` total schema labels in the released benchmark
+
+The accompanying paper reports:
+
+- Stage 1 expert-overlap validation after reconciliation: `83.72%`
+- Stage 2 strict all-annotator exact-match label reliability: `92.38%`
+- Human-human agreement on the semantic-alignment validation set: `88.86%`
+- Human-judge agreement for the selected semantic judge: `87.76%`
+
+### Who are the source data producers?
+
+The story texts are sourced from ToMBench. The belief structures and schema labels are benchmark annotations produced through the OmniToM human-validated annotation pipeline described in the accompanying paper.
+
+### Personal and Sensitive Information
+
+The release consists of short benchmark stories and belief annotations. It is not designed to contain personal user data, private communications, direct identifiers, medical records, or real-world sensitive records. As with many story-based datasets, names, family roles, occupations, emotions, intentions, and social situations may appear in the source material, but the benchmark is intended for research evaluation rather than identification or profiling.
+
+## Bias, Risks, and Limitations
+
+OmniToM is story-based, text-only, English-language, and sourced from a specific upstream benchmark distribution. It therefore reflects the representational biases, writing conventions, scenario distribution, and coverage limitations of its source stories. Names, roles, social situations, and pragmatic conventions may also be unevenly distributed across categories.
+
+Additional known limitations:
+
+- the benchmark evaluates story-based Theory of Mind rather than interactive, embodied, or multimodal social reasoning
+- the retained stories are short and self-contained, and do not stress-test long-horizon information tracking, dense temporal structure, or deeply nested mental states beyond the order-3 schema
+- the released labels come from a human-validated LLM-assisted pipeline rather than fully manual annotation of every story
+- the seven-dimensional schema is human-labeled and may retain interpretive subjectivity in socially ambiguous cases
+- Stage 1 extraction evaluation in the paper relies on a human-validated semantic judge rather than full human adjudication across the full release
+- the selected semantic judge reached `87.76%` agreement with human semantic-alignment decisions, so extraction `F1` should be interpreted as an approximate aggregate metric rather than an exact belief-level alignment score
+
+### Recommendations
+
+Users should interpret OmniToM as a diagnostic benchmark for explicit belief-structure modeling. Benchmark scores should not be treated as evidence of robust real-world interpersonal reasoning, embodied social competence, clinical or educational validity, or deployment safety.
+
+## Citation
+
+```bibtex
+@misc{omnitom2026,
+  title={OmniToM: Benchmarking Theory of Mind in LLMs via Explicit Belief Modeling},
+  author={Bawatneh, Adam and Sapkota, Sagar and Bedi, Amrit Singh and Karmaker, Santu and Shah, Mubarak},
+  year={2026},
+  eprint={2605.26322},
+  archivePrefix={arXiv}
+}
+```
+
+## More Information
+
+Paper: [OmniToM on arXiv](https://arxiv.org/abs/2605.26322)
+
+Code and evaluation scripts: [Adam-12-0/omnitom-benchmark](https://github.com/Adam-12-0/omnitom-benchmark)
+
+Project page: [OmniToM](https://adam-12-0.github.io/omnitom-project/)

croissant_rai_omnitom_corrected.json

@@ -0,0 +1,485 @@
+{
+  "@context": {
+    "@language": "en",
+    "@vocab": "https://schema.org/",
+    "arrayShape": "cr:arrayShape",
+    "citeAs": "cr:citeAs",
+    "column": "cr:column",
+    "conformsTo": "dct:conformsTo",
+    "containedIn": "cr:containedIn",
+    "cr": "http://mlcommons.org/croissant/",
+    "data": {
+      "@id": "cr:data",
+      "@type": "@json"
+    },
+    "dataBiases": "cr:dataBiases",
+    "dataCollection": "cr:dataCollection",
+    "dataType": {
+      "@id": "cr:dataType",
+      "@type": "@vocab"
+    },
+    "dct": "http://purl.org/dc/terms/",
+    "extract": "cr:extract",
+    "field": "cr:field",
+    "fileProperty": "cr:fileProperty",
+    "fileObject": "cr:fileObject",
+    "fileSet": "cr:fileSet",
+    "format": "cr:format",
+    "includes": "cr:includes",
+    "isArray": "cr:isArray",
+    "isLiveDataset": "cr:isLiveDataset",
+    "jsonPath": "cr:jsonPath",
+    "key": "cr:key",
+    "md5": "cr:md5",
+    "parentField": "cr:parentField",
+    "path": "cr:path",
+    "personalSensitiveInformation": "cr:personalSensitiveInformation",
+    "recordSet": "cr:recordSet",
+    "references": "cr:references",
+    "regex": "cr:regex",
+    "repeated": "cr:repeated",
+    "replace": "cr:replace",
+    "sc": "https://schema.org/",
+    "separator": "cr:separator",
+    "source": "cr:source",
+    "subField": "cr:subField",
+    "transform": "cr:transform",
+    "rai": "http://mlcommons.org/croissant/RAI/",
+    "prov": "http://www.w3.org/ns/prov#"
+  },
+  "@type": "sc:Dataset",
+  "distribution": [
+    {
+      "@type": "cr:FileObject",
+      "@id": "repo",
+      "name": "repo",
+      "description": "The Hugging Face git repository.",
+      "contentUrl": "https://huggingface.co/datasets/Adam-12-0/omnitom-benchmark/tree/refs%2Fconvert%2Fparquet",
+      "encodingFormat": "git+https",
+      "sha256": "https://github.com/mlcommons/croissant/issues/80"
+    },
+    {
+      "@type": "cr:FileSet",
+      "@id": "parquet-files-for-config-default",
+      "containedIn": {
+        "@id": "repo"
+      },
+      "encodingFormat": "application/x-parquet",
+      "includes": "default/*/*.parquet"
+    }
+  ],
+  "recordSet": [
+    {
+      "@type": "cr:RecordSet",
+      "dataType": "cr:Split",
+      "key": {
+        "@id": "default_splits/split_name"
+      },
+      "@id": "default_splits",
+      "name": "default_splits",
+      "description": "Splits for the default config. The released benchmark split is exposed as the Hugging Face default train split for compatibility with the dataset viewer.",
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "default_splits/split_name",
+          "dataType": "sc:Text"
+        }
+      ],
+      "data": [
+        {
+          "default_splits/split_name": "train"
+        }
+      ]
+    },
+    {
+      "@type": "cr:RecordSet",
+      "@id": "default",
+      "description": "OmniToM benchmark split exposed through the Hugging Face 'default/train' converted Parquet view. Each record is one story with a list of belief propositions and schema labels.",
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "default/split",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "fileProperty": "fullpath"
+            },
+            "transform": {
+              "regex": "default/(?:partial-)?(train)/.+parquet$"
+            }
+          },
+          "references": {
+            "field": {
+              "@id": "default_splits/split_name"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/story_id",
+          "dataType": "cr:Int64",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "story_id"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/story_category",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "story_category"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/story",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "story"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/beliefs",
+          "subField": [
+            {
+              "@type": "cr:Field",
+              "@id": "default/beliefs/actor",
+              "dataType": "sc:Text",
+              "source": {
+                "fileSet": {
+                  "@id": "parquet-files-for-config-default"
+                },
+                "extract": {
+                  "column": "beliefs"
+                },
+                "transform": {
+                  "jsonPath": "actor"
+                }
+              }
+            },
+            {
+              "@type": "cr:Field",
+              "@id": "default/beliefs/belief",
+              "dataType": "sc:Text",
+              "source": {
+                "fileSet": {
+                  "@id": "parquet-files-for-config-default"
+                },
+                "extract": {
+                  "column": "beliefs"
+                },
+                "transform": {
+                  "jsonPath": "belief"
+                }
+              }
+            },
+            {
+              "@type": "cr:Field",
+              "@id": "default/beliefs/labels",
+              "subField": [
+                {
+                  "@type": "cr:Field",
+                  "@id": "default/beliefs/labels/order",
+                  "dataType": "sc:Text",
+                  "source": {
+                    "fileSet": {
+                      "@id": "parquet-files-for-config-default"
+                    },
+                    "extract": {
+                      "column": "beliefs"
+                    },
+                    "transform": [
+                      {
+                        "jsonPath": "labels"
+                      },
+                      {
+                        "jsonPath": "order"
+                      }
+                    ]
+                  }
+                },
+                {
+                  "@type": "cr:Field",
+                  "@id": "default/beliefs/labels/truth_status",
+                  "dataType": "sc:Text",
+                  "source": {
+                    "fileSet": {
+                      "@id": "parquet-files-for-config-default"
+                    },
+                    "extract": {
+                      "column": "beliefs"
+                    },
+                    "transform": [
+                      {
+                        "jsonPath": "labels"
+                      },
+                      {
+                        "jsonPath": "truth_status"
+                      }
+                    ]
+                  }
+                },
+                {
+                  "@type": "cr:Field",
+                  "@id": "default/beliefs/labels/knowledge_access",
+                  "dataType": "sc:Text",
+                  "source": {
+                    "fileSet": {
+                      "@id": "parquet-files-for-config-default"
+                    },
+                    "extract": {
+                      "column": "beliefs"
+                    },
+                    "transform": [
+                      {
+                        "jsonPath": "labels"
+                      },
+                      {
+                        "jsonPath": "knowledge_access"
+                      }
+                    ]
+                  }
+                },
+                {
+                  "@type": "cr:Field",
+                  "@id": "default/beliefs/labels/representation",
+                  "dataType": "sc:Text",
+                  "source": {
+                    "fileSet": {
+                      "@id": "parquet-files-for-config-default"
+                    },
+                    "extract": {
+                      "column": "beliefs"
+                    },
+                    "transform": [
+                      {
+                        "jsonPath": "labels"
+                      },
+                      {
+                        "jsonPath": "representation"
+                      }
+                    ]
+                  }
+                },
+                {
+                  "@type": "cr:Field",
+                  "@id": "default/beliefs/labels/content_type",
+                  "dataType": "sc:Text",
+                  "source": {
+                    "fileSet": {
+                      "@id": "parquet-files-for-config-default"
+                    },
+                    "extract": {
+                      "column": "beliefs"
+                    },
+                    "transform": [
+                      {
+                        "jsonPath": "labels"
+                      },
+                      {
+                        "jsonPath": "content_type"
+                      }
+                    ]
+                  }
+                },
+                {
+                  "@type": "cr:Field",
+                  "@id": "default/beliefs/labels/mental_source",
+                  "dataType": "sc:Text",
+                  "source": {
+                    "fileSet": {
+                      "@id": "parquet-files-for-config-default"
+                    },
+                    "extract": {
+                      "column": "beliefs"
+                    },
+                    "transform": [
+                      {
+                        "jsonPath": "labels"
+                      },
+                      {
+                        "jsonPath": "mental_source"
+                      }
+                    ]
+                  }
+                },
+                {
+                  "@type": "cr:Field",
+                  "@id": "default/beliefs/labels/context",
+                  "dataType": "sc:Text",
+                  "source": {
+                    "fileSet": {
+                      "@id": "parquet-files-for-config-default"
+                    },
+                    "extract": {
+                      "column": "beliefs"
+                    },
+                    "transform": [
+                      {
+                        "jsonPath": "labels"
+                      },
+                      {
+                        "jsonPath": "context"
+                      }
+                    ]
+                  }
+                }
+              ]
+            }
+          ],
+          "isArray": true,
+          "arrayShape": "-1"
+        }
+      ]
+    }
+  ],
+  "conformsTo": "http://mlcommons.org/croissant/1.1",
+  "name": "omnitom-benchmark",
+  "description": "OmniToM is a benchmark for evaluating Theory of Mind in language models through explicit belief-structure modeling. Each example is a short English social-reasoning story paired with actor-centered belief propositions and seven closed-set schema labels. The release contains the 895-story benchmark split with 22,343 labeled belief propositions.",
+  "alternateName": [
+    "OmniToM",
+    "Adam-12-0/omnitom-benchmark"
+  ],
+  "creator": {
+    "@type": "Organization",
+    "name": "OmniToM authors",
+    "url": "https://huggingface.co/Adam-12-0"
+  },
+  "keywords": [
+    "text-generation",
+    "text-classification",
+    "text2text-generation",
+    "multi-class-classification",
+    "English",
+    "mit",
+    "n<1K",
+    "theory-of-mind",
+    "benchmark",
+    "social-reasoning",
+    "belief-modeling",
+    "llm-evaluation",
+    "cognitive-reasoning",
+    "mental-state-representation"
+  ],
+  "license": "https://spdx.org/licenses/MIT.html",
+  "url": "https://huggingface.co/datasets/Adam-12-0/omnitom-benchmark",
+  "rai:dataLimitations": [
+    "OmniToM is limited to English, text-only, short, self-contained Theory-of-Mind stories from seven retained ToMBench categories. It does not evaluate embodied, interactive, multimodal, long-horizon, or high-stakes social reasoning, and it should not be used as evidence of deployment-ready interpersonal reasoning or agent safety.",
+    "The benchmark contains 895 stories and 22,343 belief propositions. The retained categories and short-story format bound the phenomena covered and exclude other forms of social reasoning.",
+    "The belief structures and labels were produced by a human-validated LLM-assisted annotation pipeline rather than by exhaustive fully manual annotation of every released story. The seven-dimensional schema is human-labeled and may retain interpretive subjectivity in socially ambiguous cases, though the closed label space, three-annotator verification, majority voting, and expert adjudication reduce this risk.",
+    "Stage 1 belief-extraction evaluation in the accompanying paper uses a human-validated semantic judge rather than exhaustive human adjudication over the full benchmark. The selected judge reached 87.76% agreement with human semantic-alignment decisions, so extraction F1 should be interpreted as an approximate aggregate metric rather than an exact belief-level alignment score.",
+    "Recommended use is benchmark evaluation and diagnostic analysis. Out-of-scope uses include safety certification, clinical or educational assessment of social cognition, measurement of real-world social intelligence, profiling or decision-making about people, and treating the benchmark as a complete coverage test for all Theory-of-Mind phenomena."
+  ],
+  "rai:dataBiases": [
+    "The source stories are inherited from ToMBench, so OmniToM inherits the representational biases, writing conventions, topical coverage limits, and scenario distribution of that upstream corpus.",
+    "The benchmark retains only seven story categories judged to provide sufficiently self-contained mental-state evidence. This creates selection bias toward short, benchmark-style social-reasoning narratives and away from dialogue, multimodal evidence, culturally variable interaction patterns, and more open-ended real-world contexts.",
+    "Names, roles, cultures, social situations, pragmatic conventions, and story categories may be unevenly represented. Model performance may therefore reflect familiarity with those narrative conventions rather than general Theory-of-Mind competence across populations or contexts.",
+    "Annotations were created with LLM assistance and human validation. Although the pipeline was validated and audited, the annotation model and semantic judge may introduce model-specific biases, especially in implicit pragmatic inference, shared-knowledge judgments, and higher-order belief attribution."
+  ],
+  "rai:personalSensitiveInformation": "The dataset is not designed to contain real personal data, private communications, direct identifiers, medical records, political or religious records, or other real-world sensitive records. It consists of short benchmark stories and belief annotations. Fictional or benchmark-story names, family roles, occupations, emotions, intentions, social relationships, and social situations may appear because they are necessary for Theory-of-Mind evaluation. The dataset is English-language and should not be used for identification, profiling, or inference about real people.",
+  "rai:dataUseCases": [
+    "Validated intended use: benchmark evaluation of language models on explicit belief-structure modeling from short English social-reasoning stories. The two primary tasks are Stage 1 belief extraction and Stage 2 belief labeling under the seven-dimensional OmniToM schema.",
+    "Validated diagnostic analyses include zero-shot belief extraction, zero-shot belief labeling, semantic-judge evaluation of extracted belief tables, and analysis of multi-actor and higher-order Theory-of-Mind reasoning. The accompanying paper reports reliability evidence including 83.72% Stage 1 expert-overlap validation after reconciliation, 92.38% strict all-annotator exact-match Stage 2 label reliability, 88.86% human-human semantic-alignment agreement, model evaluation, and semantic-judge validation.",
+    "The construct represented is controlled, text-based Theory-of-Mind reasoning as explicit actor-centered belief propositions and schema labels, not broad real-world social intelligence. The benchmark is intended to make intermediate mental-state representations inspectable rather than to replace endpoint QA benchmarks entirely.",
+    "Not validated for fine-tuning as a general social-reasoning corpus, training deployed assistants, human psychological assessment, clinical diagnosis, educational placement, safety certification, fairness auditing across demographic groups, or decisions about individuals or communities."
+  ],
+  "rai:dataSocialImpact": "Positive impact: OmniToM can improve transparency in social-reasoning evaluation by moving beyond endpoint answers to explicit belief structures, helping researchers analyze failures in actor tracking, information access, source attribution, implicit inference, and higher-order belief modeling. Negative impact: scores could be misused as evidence that a model has robust real-world social intelligence, interpersonal competence, or deployment safety. Because the benchmark is story-based, English-only, and derived from a bounded upstream corpus, such over-interpretation could encourage premature deployment or mask failures in culturally diverse, interactive, multimodal, or high-stakes contexts. Mitigations include documenting limitations, recommending evaluation rather than training use, preserving upstream attribution and license information, reporting validation and judge limitations, and explicitly warning users not to treat benchmark performance as safety certification.",
+  "rai:hasSyntheticData": true,
+  "prov:wasDerivedFrom": [
+    {
+      "@id": "https://github.com/zhchen18/ToMBench",
+      "prov:label": "ToMBench",
+      "sc:license": "https://spdx.org/licenses/MIT.html",
+      "prov:wasAttributedTo": {
+        "prov:label": "ToMBench authors"
+      }
+    },
+    {
+      "@id": "https://aclanthology.org/2024.acl-long.847/",
+      "prov:label": "ToMBench paper",
+      "prov:wasAttributedTo": {
+        "prov:label": "Chen et al. 2024"
+      }
+    }
+  ],
+  "prov:wasGeneratedBy": [
+    {
+      "@type": "prov:Activity",
+      "prov:type": {
+        "@id": "https://www.wikidata.org/wiki/Q1172378"
+      },
+      "prov:label": "ToMBench source filtering and split construction",
+      "sc:description": "OmniToM begins from the 1,383-story ToMBench corpus, retains seven high-volume categories that provide sufficiently self-contained mental-state evidence, and releases 895 stories as the benchmark split.",
+      "prov:wasAttributedTo": [
+        {
+          "@type": "prov:Agent",
+          "@id": "omnitom_research_team",
+          "prov:label": "OmniToM research team",
+          "sc:description": "Authors and domain experts responsible for source filtering, split construction, benchmark design, and release preparation."
+        }
+      ]
+    },
+    {
+      "@type": "prov:Activity",
+      "prov:type": {
+        "@id": "https://www.wikidata.org/wiki/Q109719325"
+      },
+      "prov:label": "Human-validated LLM-assisted belief extraction and schema labeling",
+      "sc:description": "The benchmark annotations were produced with a human-validated LLM-assisted pipeline. Stage 1 belief structures were reconciled by domain experts, and Stage 2 labels were verified by three annotators through majority vote or expert adjudication. The finalized annotation setup yielded 22,343 belief propositions and 156,401 schema labels across the 895-story benchmark split.",
+      "prov:wasAttributedTo": [
+        {
+          "@type": "prov:Agent",
+          "@id": "omnitom_domain_experts_and_annotators",
+          "prov:label": "OmniToM domain experts and annotators",
+          "sc:description": "Human annotators and domain experts who performed schema refinement, pilot annotation, belief extraction, schema labeling, semantic-alignment annotation, validation, adjudication, and validation."
+        }
+      ]
+    },
+    {
+      "@type": "prov:Activity",
+      "prov:type": {
+        "@id": "https://www.wikidata.org/wiki/Q3306762"
+      },
+      "prov:label": "Quality review, validation, and semantic-judge validation",
+      "sc:description": "The paper reports multiple validation checks: 83.72% Stage 1 expert-overlap validation after reconciliation, 92.38% strict all-annotator exact-match Stage 2 label reliability, 88.86% human-human semantic-alignment agreement on jointly annotated belief rows, and semantic-judge validation. GPT-5 achieved the highest reported human-judge agreement among tested judge candidates (87.76%) and was fixed as the independent Stage 1 semantic judge. These checks support benchmark use while documenting that Stage 1 extraction scores remain approximate.",
+      "prov:wasAttributedTo": [
+        {
+          "@type": "prov:Agent",
+          "@id": "omnitom_research_team",
+          "prov:label": "OmniToM research team",
+          "sc:description": "Team responsible for benchmark validation, quality review, and validation reporting."
+        },
+        {
+          "@type": "prov:SoftwareAgent",
+          "@id": "gpt_5",
+          "prov:label": "GPT-5",
+          "sc:description": "Selected as the fixed Stage 1 semantic judge after validation against human semantic-alignment decisions."
+        }
+      ]
+    }
+  ],
+  "datePublished": "2026-05-04",
+  "version": "1.0.0",
+  "inLanguage": "en",
+  "isAccessibleForFree": true,
+  "citeAs": "@article{bawatneh2026omnitom,\n  title={OmniToM: Benchmarking Theory of Mind in LLMs via Explicit Belief Modeling},\n  author={Bawatneh, Adam and Sapkota, Sagar and Bedi, Amrit Singh and Karmaker, Santu and Shah, Mubarak},\n  journal={arXiv preprint arXiv:2605.26322},\n  year={2026}\n}"
+}