GSM-Hard
GSM8K test problems with unusually large numerical values, designed to separate reasoning from easy arithmetic.
The public 1,319-item transformation remains a useful arithmetic stress test, but automatic number replacement creates implausible and sometimes invalid questions; it should not be used as a standalone reasoning benchmark.
Performance Timeline
Longitudinal progression of model scores against human baselines.Human Baseline & Difficulty Horizon
Calibrated human reference points, specialist benchmarks, and ceiling thresholds.Human calculation accuracy when numeric constants are replaced with large arbitrary numbers causing manual arithmetic errors.
Metric & Scoring Methodology
Verification protocols, aggregation formulas, and specialized metric variants.final-answer accuracy (%)Dataset & Compute Cost
Evaluation volume, public availability, API pricing, and local hardware requirements.$5 – $20 USD for full benchmark evaluation run on frontier APIs.
1x NVIDIA RTX 4090 (24GB) or A100 (40GB/80GB) via vLLM / SGLang
How to Run & Reproduce
Standardized evaluation protocols, CLI commands, and reproducible runner templates.lm_eval --model hf --model_args pretrained=<model_path> --tasks gsm-hard --batch_size autoopencompass --datasets gsm-hard --models <model_config># Standard API Evaluation Loop
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user", "content": prompt}],
temperature=0.0,
)When publishing results for GSM-Hard, always report the exact prompt template, few-shot exemplar ordering, sampling temperature (temperature=0), maximum reasoning budget tokens, and the precise timestamped model snapshot ID.
Contamination & Memorization Analysis
Audit of pretraining exposure risks, memorization vectors, and refresh policies.Static fixed snapshot
Public on web / HuggingFace