Instruction FollowingNearing Saturation
Trust Score:
80

ComplexBench

Bilingual benchmark of how models follow instructions when several constraints must hold at once — where stacking constraints is not linear.

Launched: —Refresh: static
Status Assessment (nearing-saturation):

Moved to the site's nearing-saturation bucket: ComplexBench remains useful for multi-constraint composition analysis, but it is a public static set and should be treated more as a diagnostic successor to simpler instruction-following checks than as a durable frontier-ranking benchmark.

Performance Timeline

Longitudinal progression of model scores against human baselines.
No performance score history recorded yet.

Human Baseline & Difficulty Horizon

Calibrated human reference points, specialist benchmarks, and ceiling thresholds.
Measured Human Score76.0%Domain Expert Baseline
Baseline Protocol & Interpretation

Human prompt engineers following complex multi-constraint generation instructions.

Metric & Scoring Methodology

Verification protocols, aggregation formulas, and specialized metric variants.
Primary Metric:overall constraint-satisfaction (%)
Scoring Engine:llm-judge

Dataset & Compute Cost

Evaluation volume, public availability, API pricing, and local hardware requirements.
Total Dataset Size1,150Annotated evaluation items
Public Test SetPublicOpenly mirrored on repositories
Access GatingOpen AccessUnrestricted download
Evaluation LicenseOpen AccessDataset usage and redistribution terms
Frontier API Compute Cost:

$5 – $20 USD for full benchmark evaluation run on frontier APIs.

Recommended Local GPU Setup:

1x NVIDIA RTX 4090 (24GB) or A100 (40GB/80GB) via vLLM / SGLang

Official Dataset & Benchmark Files:Download / View Dataset Repository ↗

How to Run & Reproduce

Standardized evaluation protocols, CLI commands, and reproducible runner templates.
Prompt Regimezero-shot
Reasoning Modedirect
Sampling Temp0
Pass@k Budgetk = 1
Tools & SandboxPure Text
Scoring Verifierllm-judge
Option AEleutherAI LM-Evaluation-Harness (Open-Weight Models)
lm_eval --model hf --model_args pretrained=<model_path> --tasks complexbench --batch_size auto
Option BOpenCompass Evaluation Framework
opencompass --datasets complexbench --models <model_config>
Python APIDeterministic Inference Loop Snippet
# 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,
)
Standardized Reporting Requirement:

When publishing results for ComplexBench, 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.
Overall Contamination Risk:LOW
Refresh Cadence:

Static fixed snapshot

Test Set Exposure:

Public on web / HuggingFace