Contamination MapText / LLM / Model
Phi-4-mini
- Organisation
- Microsoft
- Family
- Phi · version 4-mini
- Released
- 2025-02-19 (HF repo creation of microsoft/Phi-4-mini-instruct (2025-02-19); card: 'Release date: February 2025')
- Training cutoff
- 2024-06-01
Evidence
“This is a static model trained on offline datasets with the cutoff date of June 2024 for publicly available data.”
huggingface.co · model card · Status · retrieved 2026-10-15 - Checkpoints
- microsoft/Phi-4-mini-instruct
- Weights
- Open · MIT
- Links
- PaperModel card
Training data disclosure: Vague
Level determined for 6% of benchmark cells.
- TestTest data seen0
- SeenSeen, split unclear0
- TrainTraining data seen0
- Related data seen0
- NoNot seen1
- ?Unknown14
“With these techniques, we built the 5 trillion pre-training data corpus, which is larger and in higher quality compared to the Phi-3.5-Mini.”
Other declared training data
-
publicly available documents filtered for quality, educational data and code (broad web data: may contain benchmark test sets) · split: unknown
Evidence
“publicly available documents filtered for quality, selected high-quality educational data, and code”
huggingface.co · model card · Training Datasets · retrieved 2026-10-15 -
Phi-4 synthetic data (not yet matched to a dataset) · split: unknown
Evidence
“we incorporated Phi-4 synthetic data [AAB+ 24] into this model training with the same processing and decontamination.”
arxiv.org · paper · Section 3.1.1 · retrieved 2026-10-15 -
instruction-based math and coding data (not yet matched to a dataset) · split: unknown
Evidence
“For the math and coding data, we have augmented our original data with a specific instruction-based math and coding data set.”
arxiv.org · paper · Section 3.1.1 · retrieved 2026-10-15 -
post-training data (function calling, summarization, synthetic instruction-following, code completion) (not yet matched to a dataset) · split: unknown · fine-tuning
Evidence
“Phi-4-Mini includes a significantly larger and more diverse set of function calling and summarization data.”
arxiv.org · paper · Section 3.1.2 · retrieved 2026-10-15
Benchmark cells
Classic LLM benchmarks
Unknown MMLU (Measuring Massive Multitask Language Understanding) · training data not disclosed in enough detail
Unknown GSM8K (Grade School Math 8K) · training data not disclosed in enough detail
Unknown MATH (Mathematics Aptitude Test of Heuristics) · training data not disclosed in enough detail
Unknown HumanEval · training data not disclosed in enough detail
Unknown ARC-Challenge (AI2 Reasoning Challenge, Challenge Set) · training data not disclosed in enough detail
Unknown HellaSwag · training data not disclosed in enough detail
Unknown TruthfulQA · training data not disclosed in enough detail
Unknown WinoGrande · training data not disclosed in enough detail
Recent LLM benchmarks
Unknown GPQA Diamond · training data not disclosed in enough detail
Unknown MMLU-Pro · training data not disclosed in enough detail
Unknown LiveCodeBench (code generation, release_v6) · training data not disclosed in enough detail
Unknown SWE-bench Verified · training data not disclosed in enough detail
Unknown AIME 2024 · training data not disclosed in enough detail
Not seen AIME 2025 · its test items were published from 2025-02-06, after the model's stated training-data cutoff (2024-06-01)
Evidence (3)
“This is a static model trained on offline datasets with the cutoff date of June 2024 for publicly available data.”
“2025 AIME I problems and solutions. The test was held on Thursday, February 6, 2025.”
“This year the AIME I will take place on February 6, 2025, while the alternate AIME II will take place on February 12, 2025.”
Unknown MATH-500 · training data not disclosed in enough detail
Notes
The card describes an n-gram decontamination step ('Samples with matching n-grams above a threshold were flagged as contaminated and removed') but does not name the benchmarks, so no exclusion is recorded. Phi-4-multimodal-instruct (built on Phi-4-mini with LoRA adapters) is not covered.
Sources
- Phi-4-Mini Technical Report · paper · retrieved 2026-10-15
- Phi-4-mini-instruct model card · model card · retrieved 2026-10-15