WildChat vs The Stack v2

Aggregates over 1 million authentic user conversations with ChatGPT and GPT-4 captured through a research chatbot interface, preserving full conversation threads with metadata including timestamps, user demographics (country, browser type), and conversation-level toxicity annotations. The dataset captures genuine, unfiltered user intents across diverse domains without synthetic generation or prompt engineering, enabling analysis of actual AI usage patterns in production environments.

Unique: Captures unfiltered, real-world conversations from production ChatGPT/GPT-4 deployments rather than synthetic or crowdsourced data, preserving authentic user intents, failure modes, and edge cases with demographic metadata (country, browser) enabling stratified analysis across user populations

vs alternatives: Larger scale (1M+ conversations) and more authentic than crowdsourced datasets like ShareGPT, with explicit demographic metadata absent from most open conversation corpora, though less curated and safety-filtered than instruction-tuning datasets like FLAN or Alpaca

Enables filtering and analysis of conversations by user demographics (country, browser type) and conversation-level metadata, allowing researchers to slice the dataset by geographic region, device type, or other user attributes. The dataset structure preserves demographic fields as queryable attributes, supporting cohort analysis, geographic bias detection, and population-specific model evaluation without requiring external demographic inference.

Unique: Provides explicit demographic metadata (country, browser) at conversation level, enabling direct stratified analysis without requiring external demographic inference or proxy models, though limited to coarse-grained attributes compared to crowdsourced alternatives

vs alternatives: More direct demographic stratification than ShareGPT or other conversation corpora, though less granular than purpose-built fairness datasets with rich demographic annotations

Provides conversation-level toxicity labels assigned through automated or human annotation, enabling researchers to identify and filter harmful content, study safety patterns, and train content moderation models. Labels are attached at the conversation level (not per-message), allowing downstream filtering of unsafe conversations or stratified analysis of toxicity distribution across user demographics and conversation types.

Unique: Provides real-world toxicity annotations from production ChatGPT/GPT-4 conversations rather than synthetic or crowdsourced toxic examples, capturing authentic harmful content patterns without artificial prompt engineering, though at conversation-level granularity rather than message-level

vs alternatives: More authentic toxicity examples than synthetic safety datasets, though coarser-grained labeling and less detailed harm taxonomy than purpose-built safety datasets like ToxiGen or RealToxicityPrompts

Provides access to non-English conversations within the dataset, enabling analysis of how users in different languages interact with English-trained LLMs and supporting training of multilingual or cross-lingual models. Conversations are preserved in original language with metadata indicating language or country of origin, allowing language-specific filtering and comparative analysis across linguistic communities.

Unique: Includes real-world multilingual conversations from production ChatGPT/GPT-4 deployments, capturing authentic non-English user interactions and code-switching patterns, though limited in coverage and requiring language detection for explicit language identification

vs alternatives: More authentic multilingual examples than synthetic multilingual datasets, though smaller and less balanced than purpose-built multilingual corpora like FLORES or mC4

Provides structured metadata for each conversation including timestamps, conversation IDs, user IDs, and conversation length, enabling temporal analysis of usage patterns, trend detection, and time-series studies of how user needs and LLM interactions evolved. Metadata is queryable and filterable, supporting cohort analysis by time period and correlation analysis between temporal patterns and conversation characteristics.

Unique: Preserves conversation-level timestamps from production ChatGPT/GPT-4 deployments, enabling temporal analysis of real-world usage evolution without synthetic time-shifting, though limited to conversation-level granularity without turn-level timing

vs alternatives: More authentic temporal data than synthetic datasets, though coarser-grained than specialized time-series conversation corpora with explicit turn-level timestamps

Provides conversations spanning diverse user intents and domains (coding help, creative writing, sensitive topics, general Q&A, etc.) captured from real users without prompt engineering, enabling researchers to sample representative conversations across use cases and train models on realistic domain distributions. The dataset's scale and authenticity allow stratified sampling by inferred domain or use case without requiring explicit domain labels.

Unique: Captures authentic domain diversity from real ChatGPT/GPT-4 users without synthetic prompt engineering, preserving natural distribution of use cases and user intents, though requiring post-hoc domain inference rather than explicit labels

vs alternatives: More authentic domain diversity than synthetic instruction-tuning datasets, though less explicitly labeled and curated than purpose-built domain-specific corpora

The dataset includes structured metadata for each conversation (user demographics, browser/device info, conversation length, timestamps, toxicity labels) that can be extracted and aggregated for statistical analysis. Researchers can compute summary statistics (e.g., average conversation length by country, toxicity prevalence by domain) without processing full conversation text, enabling efficient exploratory analysis and dataset characterization. Metadata is stored in queryable fields, supporting both individual record lookup and bulk aggregation.

Unique: Provides structured metadata fields (country, browser, device, toxicity label) linked to each conversation, enabling efficient statistical summarization without processing full conversation text. Metadata is captured at collection time, preserving temporal and contextual information.

vs alternatives: More efficient for statistical analysis than processing full conversation text, but metadata quality and completeness are not explicitly documented compared to explicitly validated datasets

The dataset captures authentic user requests and model responses, enabling analysis of instruction-following patterns, user intent distribution, and how well models address diverse user needs. Researchers can analyze which types of instructions users provide, how models interpret and respond to them, and where misalignment or misunderstanding occurs. This supports studying instruction-following quality, identifying common user frustrations, and understanding the diversity of real-world use cases beyond typical benchmarks.

Unique: Captures authentic user instructions and model responses from production ChatGPT/GPT-4 deployments, reflecting real instruction-following challenges and user intent distribution rather than synthetic instruction-tuning data. Includes edge cases and sensitive topics that users genuinely request.

vs alternatives: More representative of real-world instruction-following patterns than synthetic instruction-tuning datasets, but lacks explicit success metrics or user satisfaction labels compared to explicitly validated instruction-following benchmarks

Aggregates 67 TB of source code from the Software Heritage archive, filtering for permissively licensed repositories (MIT, Apache 2.0, BSD, etc.) across 600+ programming languages. Uses automated license detection and validation to ensure legal compliance for model training. Implements a rigorous deduplication pipeline at file and repository levels to eliminate redundant training data and reduce dataset bloat.

Unique: Largest open-source code dataset at 67 TB with automated opt-out governance allowing repository owners to request removal, combined with rigorous deduplication and PII removal pipeline — no other public dataset offers this scale with legal compliance and community control mechanisms

vs alternatives: Larger and more legally compliant than GitHub's CodeSearchNet (14M files) or Google's BigQuery public datasets, with explicit opt-out governance vs. implicit inclusion, and covers 600+ languages vs. Codex training data's undisclosed language distribution

Implements a community-driven opt-out system where repository owners can request removal of their code from the dataset without legal takedown notices. Maintains a registry of excluded repositories and re-applies exclusions during dataset updates. Provides transparent governance documentation and a clear submission process for removal requests, balancing open access with creator rights.

Unique: First large-scale code dataset to implement opt-out governance at dataset level rather than relying solely on license compliance, with transparent registry and community submission process — shifts power from dataset creators to code contributors

vs alternatives: More respectful of creator autonomy than GitHub Copilot's training approach (no opt-out) or academic datasets (one-time snapshot), and more scalable than individual DMCA takedowns

Automated pipeline that scans source code for personally identifiable information (email addresses, API keys, SSH keys, credit card patterns, phone numbers) and removes or redacts them before dataset release. Uses regex patterns, entropy-based detection for secrets, and heuristic rules to identify sensitive data. Operates at file level with configurable sensitivity thresholds to balance data utility against privacy risk.

Unique: Combines regex pattern matching, entropy-based secret detection, and heuristic rules in a unified pipeline with configurable sensitivity — more comprehensive than simple regex-only approaches, but trades off false positive rate against security coverage

vs alternatives: More thorough than GitHub's secret scanning (which only flags known patterns) because it includes entropy-based detection for unknown secret formats, but less accurate than specialized tools like TruffleHog due to language-agnostic approach

Indexes 67 TB of source code across 600+ programming languages with language-aware metadata (syntax, file extension, language family). Enables retrieval by language, license, repository, or code patterns. Uses Software Heritage's existing indexing infrastructure as foundation, augmented with language detection and classification. Supports both bulk download and filtered queries for specific language subsets.

Unique: Leverages Software Heritage's existing language detection and indexing infrastructure, then augments with BigCode-specific language classification and filtering — avoids reinventing language detection while providing dataset-specific query capabilities

vs alternatives: More comprehensive language coverage (600+ languages) than GitHub's Linguist (500+ languages) and more accessible than Software Heritage's raw API because it's pre-filtered for permissive licenses and deduplicated

Removes duplicate code files and repositories using content hashing (SHA-256 or similar) and fuzzy matching for near-duplicates. Operates in two stages: exact deduplication via hash matching, then fuzzy matching (e.g., Jaccard similarity or MinHash) to catch semantically identical code with minor formatting differences. Preserves one canonical copy of each unique code pattern while removing redundant training examples.

Unique: Two-stage deduplication combining exact hash matching with fuzzy similarity matching (likely MinHash or Jaccard) to catch both identical and near-identical code — more thorough than single-stage approaches but computationally expensive

vs alternatives: More aggressive deduplication than CodeSearchNet (which uses simple hash matching) because it catches near-duplicates, but less semantic than clone detection tools (which understand code structure) because it's content-based

Integrates with Software Heritage's comprehensive archive of 200+ million repositories and their full version control history. Extracts source code snapshots from Software Heritage's Git/Mercurial/SVN repositories, preserving repository metadata (commit history, author info, timestamps). Provides access to code at specific points in time, enabling historical analysis or training on code evolution patterns.

Unique: Leverages Software Heritage's universal code archive (200M+ repositories) as data source, providing access to code that would be impossible to collect via GitHub API alone — enables training on archived/deleted repositories and non-GitHub platforms (GitLab, Gitea, etc.)

vs alternatives: More comprehensive than GitHub-only datasets because it includes code from GitLab, Gitea, SourceForge, and other platforms archived by Software Heritage; more legally defensible than web scraping because it uses an established, community-maintained archive

Tracks and validates SPDX license identifiers for each repository, ensuring only permissively licensed code (MIT, Apache 2.0, BSD, etc.) is included. Maintains license metadata alongside code files, enabling downstream users to verify legal compliance. Implements license hierarchy and compatibility checking to handle dual-licensed or complex licensing scenarios.

Unique: Combines automated SPDX detection with manual review and maintains license metadata alongside code, enabling downstream users to verify compliance — more transparent than datasets that simply claim 'permissive licenses' without proof

vs alternatives: More legally rigorous than GitHub's CodeSearchNet (which doesn't validate licenses) and more transparent than Codex training data (which doesn't disclose license filtering at all)

Maintains versioned snapshots of the dataset (e.g., v2.0, v2.1) with documented changes between versions (new repositories added, deduplication improvements, PII removal updates). Provides checksums and manifests for reproducibility, enabling researchers to cite specific dataset versions and reproduce results. Tracks dataset lineage and transformation history.

Unique: Maintains semantic versioning and detailed changelogs for dataset releases, enabling researchers to cite specific versions and understand dataset evolution — more rigorous than one-off dataset releases without versioning

vs alternatives: More reproducible than academic datasets that are released once without versioning, and more transparent than commercial datasets (Codex) that don't disclose version history or changes