metrics.json

{
  "model": "Helios-Reasoner-7B",
  "checkpoint": "step_1200",
  "overall_eval_accuracy": 0.725,
  "num_benchmarks": 15,
  "benchmarks": [
    {"name": "code_generation", "score": 0.6735},
    {"name": "common_sense", "score": 0.75},
    {"name": "creative_writing", "score": 0.6357},
    {"name": "dialogue_generation", "score": 0.6602},
    {"name": "instruction_following", "score": 0.77},
    {"name": "knowledge_retrieval", "score": 0.6879},
    {"name": "logical_reasoning", "score": 0.8383},
    {"name": "math_reasoning", "score": 0.5727},
    {"name": "question_answering", "score": 0.6179},
    {"name": "reading_comprehension", "score": 0.7176},
    {"name": "safety_evaluation", "score": 0.75},
    {"name": "sentiment_analysis", "score": 0.8},
    {"name": "summarization", "score": 0.7786},
    {"name": "text_classification", "score": 0.8377},
    {"name": "translation", "score": 0.8111}
  ]
}