chinese-llm-benchmark vs xCodeEval

Evaluates Chinese LLMs across 8 major domains (Medical, Education, Finance, Law, Administrative Affairs, Psychological Health, Reasoning & Math, Language & Instruction Following) using approximately 300 specific evaluation dimensions. Each domain assessment aggregates task-specific scores (1-5 scale per question) normalized to 0-100 point scale, then combines domain scores to produce overall model rankings. The framework uses domain-specific test questions designed to measure real-world capability rather than general language understanding.

Unique: Combines 8 specialized domain evaluations (Medical, Finance, Law, etc.) with ~300 evaluation dimensions specifically designed for Chinese LLMs, rather than generic language benchmarks. Aggregates individual question scores (1-5 scale) into normalized domain scores (0-100) then composite rankings, enabling cross-domain capability comparison. Maintains 2M+ defect library linking model failures to specific domains for root-cause analysis.

vs alternatives: Deeper domain specialization than MMLU or C-Eval (which focus on general knowledge) and Chinese-specific evaluation design vs English-centric benchmarks like HELM or LMSys Chatbot Arena

Organizes 298 evaluated models into hierarchical leaderboards using primary classification (commercial vs open-source) and secondary tiers (price tier for commercial models, parameter size for open-source models). The system maintains separate ranked lists for each category, enabling users to compare models within similar cost/capability profiles. Leaderboard data is stored in markdown files (commerce2.md, reasonmodel.md, alldata.md) with model metadata (name, version, provider, parameters, pricing) and performance scores aggregated from domain evaluations.

Unique: Implements multi-dimensional leaderboard organization (commercial/open-source primary split, then price tier or parameter size secondary split) with separate ranked lists for reasoning-specialized models. Uses markdown-based leaderboard storage (commerce2.md, reasonmodel.md, alldata.md) enabling version control and community contributions. Maintains model metadata (provider, parameters, pricing) alongside evaluation scores for context-aware comparison.

vs alternatives: More granular category-based filtering than MMLU leaderboards (which use single global ranking) and explicit price-tier organization vs Hugging Face Model Hub (which lacks domain-specific performance context)

Maintains comprehensive metadata for 298+ evaluated models including name, version, provider/developer organization, model type (commercial/open-source), parameter count, pricing information, release date, and availability status. Metadata is stored alongside evaluation scores in leaderboard files and enables filtering, sorting, and comparison based on model attributes. The system tracks model evolution (versions, updates) and maintains historical metadata for deprecated or superseded models.

Unique: Maintains comprehensive metadata for 298+ models (name, version, provider, parameters, pricing, availability) alongside evaluation scores in leaderboard files. Enables attribute-based filtering and comparison (by provider, parameter size, pricing tier). Tracks model versions and evolution over time within version-controlled repository.

vs alternatives: Integrated metadata with evaluation scores vs separate model registries (Hugging Face, OpenRouter) and version-controlled metadata history vs static model information

Maintains a defect library containing over 2 million documented model errors collected during evaluation across all domains and models. The system indexes failures by model, domain, question type, and error category, enabling researchers to identify systematic failure patterns. Defect records link specific model errors to evaluation questions, domain context, and error classification, supporting root-cause analysis and model improvement research. The library serves as a queryable knowledge base for understanding model weaknesses rather than just performance scores.

Unique: Aggregates 2M+ model failures into indexed defect library linked to specific evaluation questions, domains, and models — enabling systematic error pattern analysis rather than just aggregate scores. Supports cross-model error comparison to identify shared weaknesses and domain-specific failure distributions. Provides raw failure examples for fine-tuning and adversarial testing rather than only summary statistics.

vs alternatives: More comprehensive failure documentation than MMLU or C-Eval (which report only aggregate accuracy) and enables error-driven model improvement vs score-only benchmarks

Implements specialized evaluation for Chinese language understanding and instruction following, including Gaokao (Chinese college entrance exam) level questions that test reading comprehension, writing quality, and complex reasoning in Chinese. The evaluation framework includes domain-specific language tasks (medical terminology understanding, legal document interpretation, financial report analysis) alongside general Chinese language proficiency assessment. Scoring incorporates both accuracy and response quality (1-5 scale) to capture nuanced language performance beyond binary correctness.

Unique: Incorporates Gaokao (Chinese college entrance exam) level questions into evaluation framework, testing academic-level Chinese language understanding and writing quality. Combines general language proficiency assessment with domain-specific language tasks (medical terminology, legal documents, financial reports in Chinese). Uses 1-5 quality scale for response evaluation rather than binary correctness, capturing nuanced language performance.

vs alternatives: Chinese-specific academic assessment vs English-centric benchmarks (MMLU, HELM) and Gaokao-level difficulty calibration vs generic language benchmarks

Evaluates models on mathematical computation, logical reasoning, and complex problem-solving through domain-specific test questions in the 'Reasoning & Math' category. The evaluation framework assesses both correctness of final answers and quality of reasoning steps (1-5 scale), capturing partial credit for correct methodology with computational errors. Supports multi-step reasoning problems, symbolic manipulation, and logical inference tasks designed to test mathematical capability beyond simple arithmetic.

Unique: Evaluates mathematical reasoning with 1-5 quality scale for reasoning steps rather than binary correctness, enabling partial credit for correct methodology with computational errors. Combines final answer accuracy with reasoning quality assessment to capture mathematical thinking capability. Includes multi-step reasoning problems and logical inference tasks beyond simple arithmetic.

vs alternatives: More nuanced mathematical assessment than MMLU (binary correctness) and captures reasoning quality vs answer-only evaluation

Implements specialized evaluation across four professional domains (Medical, Finance, Law, Administrative Affairs) with domain-expert-designed test questions requiring specialized knowledge and reasoning. Each domain assessment uses realistic scenarios (medical case studies, financial analysis problems, legal document interpretation, administrative policy questions) to evaluate practical professional capability rather than general knowledge. Scoring incorporates domain-specific rubrics reflecting professional standards and best practices in each field.

Unique: Evaluates four professional domains (Medical, Finance, Law, Administrative) using domain-expert-designed test questions with realistic scenarios (medical case studies, financial analysis, legal document interpretation) rather than generic knowledge questions. Incorporates domain-specific scoring rubrics reflecting professional standards and best practices. Enables cross-domain comparison to identify models suitable for professional applications.

vs alternatives: More specialized domain assessment than general benchmarks (MMLU, C-Eval) and realistic professional scenarios vs academic knowledge questions

Evaluates models on psychological health concepts, mental health counseling knowledge, and psychological reasoning through specialized test questions in the 'Psychological Health' domain. Assessment covers mental health terminology, therapeutic approaches, psychological assessment, and ethical counseling practices. Scoring incorporates both knowledge accuracy and quality of psychological reasoning (1-5 scale) to evaluate capability for mental health support applications.

Unique: Specialized evaluation of psychological health knowledge and mental health counseling capability using domain-specific test questions. Incorporates 1-5 quality scale for psychological reasoning assessment. Addresses sensitive domain requiring both knowledge accuracy and ethical appropriateness in responses.

vs alternatives: Dedicated mental health domain assessment vs general benchmarks lacking psychological expertise, and explicit safety consideration for sensitive mental health applications

Provides a standardized evaluation framework for code generation models that accepts generated code in 17 programming languages (C, C++, C#, Java, Kotlin, Go, Rust, Python, Ruby, PHP, JavaScript, Perl, Haskell, OCaml, Scala, D, Pascal) and validates correctness through actual execution against unit tests via the ExecEval Docker-based execution engine. Uses a centralized problem definition model with src_uid foreign keys linking generated code to shared problem descriptions and unittest_db.json, enabling consistent evaluation across language variants of the same problem.

Unique: Combines 25M training examples across 7,500 unique problems with an execution-based evaluation pipeline (ExecEval) that actually runs generated code in Docker containers against unit tests, rather than relying on static analysis or string matching. The src_uid linking system creates a normalized data model where problem descriptions and tests are stored once and referenced by all language variants, eliminating duplication and ensuring consistency.

vs alternatives: Larger scale (25M examples vs typical 10-100K) and true execution-based validation across more languages (17 vs 4-6) than HumanEval or CodeXGLUE, with explicit support for code translation and repair tasks beyond generation.

Implements a foreign key linking system where all task-specific datasets (program synthesis, code translation, APR, retrieval) reference shared problem definitions via src_uid identifiers. Problem descriptions and unit tests are stored once in centralized problem_descriptions.jsonl and unittest_db.json files, then linked by src_uid to avoid duplication. The Hugging Face datasets API automatically resolves these links during data loading, returning enriched DatasetDict objects with problem context pre-joined to task examples.

Unique: Uses a normalized relational data model (src_uid as foreign key) for a code benchmark, treating problem definitions as a separate entity layer rather than embedding them in each task dataset. This is more sophisticated than typical flat-file benchmark structures and enables consistent multi-task evaluation on identical problems.

vs alternatives: More efficient than duplicating problem descriptions across 7 task datasets (reduces storage by ~30-40%), and enables automatic link resolution via Hugging Face API unlike manual CSV joins in CodeXGLUE or HumanEval variants.

Provides a Python API for loading xCodeEval datasets from Hugging Face Hub (NTU-NLP-sg/xCodeEval) with automatic src_uid-based linking between task datasets and shared problem definitions. The datasets library handles data downloading, caching, and streaming, while the xCodeEval integration automatically joins task examples with problem_descriptions.jsonl and unittest_db.json using src_uid foreign keys. Returns DatasetDict objects with enriched examples ready for model training or evaluation.

Unique: Integrates xCodeEval with Hugging Face datasets library, providing automatic src_uid resolution and streaming support. Treats data loading as a first-class concern with built-in linking logic, rather than requiring manual JSON parsing.

vs alternatives: More convenient than manual Git LFS downloads because it handles caching and automatic linking, and integrates seamlessly with Hugging Face training pipelines vs custom data loaders.

Provides an alternative data access method using Git LFS for users who prefer direct file access or need selective dataset downloads. Supports cloning the repository with LFS disabled, then pulling specific task files or problem definitions on demand. Useful for custom processing pipelines or environments where Python/Hugging Face is not available, though requires manual src_uid linking to join task examples with problem definitions.

Unique: Provides Git LFS-based alternative to Hugging Face API, enabling direct file access and selective downloads. Requires manual src_uid linking but offers more control over data access patterns.

vs alternatives: More flexible than Hugging Face API for selective downloads and custom pipelines, but requires more manual work for src_uid linking and lacks automatic caching/streaming.

Implements a standardized three-phase evaluation pipeline (Phase 1: Generation, Phase 2: Execution, Phase 3: Metrics) that applies consistently across all 7 tasks (program synthesis, code translation, APR, tag classification, code compilation, NL-code retrieval, code-code retrieval). Phase 1 generates or retrieves code, Phase 2 executes it via ExecEval or computes retrieval metrics, and Phase 3 aggregates results into pass@k, MRR, NDCG, or other task-specific metrics. Enables direct comparison of model performance across tasks.

Unique: Defines a unified three-phase evaluation pipeline that applies to all 7 tasks, treating generation, execution, and metric computation as separate concerns. Enables consistent evaluation methodology across diverse task types (generation, translation, retrieval, classification).

vs alternatives: More comprehensive than task-specific evaluation scripts because it provides a unified framework for all 7 tasks, and enables direct comparison of model performance across different task types.

Evaluates code generation models on the program synthesis task by accepting natural language problem descriptions and generating code solutions in any of 17 languages. The evaluation pipeline (Phase 1: Generation, Phase 2: Execution, Phase 3: Metrics) runs generated code against unit tests via ExecEval, computing pass@k metrics (pass@1, pass@10, etc.) that measure the probability of finding a correct solution within k samples. Supports both single-solution and multi-sample evaluation modes for assessing model reliability.

Unique: Implements a three-phase evaluation pipeline (Generation → Execution → Metrics) with explicit pass@k computation that measures the probability of finding a correct solution within k attempts, rather than just binary pass/fail. Supports multi-sample evaluation across 17 languages with language-specific compiler configurations and timeout handling.

vs alternatives: More rigorous than HumanEval's simple pass@k because it handles language-specific compilation errors and timeouts explicitly, and scales to 25M training examples vs HumanEval's 164 problems.

Evaluates code translation models by accepting source code in one language and generated translations in a target language, then validating functional equivalence through execution against shared unit tests. The translation evaluation pipeline compiles and executes both source and translated code against the same unittest_db.json test cases, comparing outputs to detect translation errors. Supports all 17 language pairs (though not all pairs may have training data) and uses language-specific compiler mappings to handle syntax differences.

Unique: Validates code translation by executing both source and target code against identical unit tests and comparing outputs, ensuring functional equivalence rather than syntactic similarity. Uses language-specific compiler mappings to handle the complexity of 17 different compilation environments and their idiosyncrasies.

vs alternatives: More rigorous than BLEU-score-based translation metrics because it validates actual functional correctness through execution, and covers more language pairs (17 vs typical 2-4) with explicit compiler integration.

Evaluates program repair models by providing buggy code snippets and expecting corrected versions that pass unit tests. The APR evaluation pipeline executes repaired code against unittest_db.json test cases, measuring whether the repair successfully fixes the bug without introducing new failures. Supports repairs across all 17 languages and uses the same execution-based validation as program synthesis, enabling direct comparison of repair quality.

Unique: Treats program repair as an executable task where success is measured by unit test passage, rather than syntactic similarity to reference repairs. Integrates with the same ExecEval pipeline as program synthesis, enabling direct performance comparison between generation and repair models.

vs alternatives: More comprehensive than traditional APR benchmarks (Defects4J, QuixBugs) because it covers 17 languages and 7,500 problems vs 395 Java bugs, and uses consistent execution-based metrics across all repair types.