Skip to content
H
Howardism
Howardism · Vol. 03Plate I · No. 01

Capability Evaluation, tagged.

Notes17TagCapability EvaluationOldest7 Jun 2026Newest15 Oct 2026

Every article tagged capability evaluation, newest first.

  1. C01
    Epoch Automation Reports
    Capability EvaluationReal Work EvaluationOpen Weights+1
    · 15′
  2. C02
    Recurring Production Agent Evaluation
    BenchmarksEvaluation MethodologyCapability Evaluation+3
    · 15′
  3. C03
    Cognitive Capability Profiling for Task Suitability
    Evaluation MethodologyPsychometricsCapability Evaluation+3
    · 37′
  4. C04
    Item Response Theory for LLM Benchmarks
    BenchmarksEvaluation MethodologyPsychometrics+2
    · 36′
  5. C05
    Adaptive Stopping in Evaluation Sampling
    Capability EvaluationBenchmarksEvaluation Methodology+3
    · 27′
  6. C06
    Covert Capabilities
    AlignmentSafetyCapability Evaluation+2
    · 24′
  7. C07
    Expenditure Horizon
    BenchmarksCapability EvaluationEvaluation Methodology+3
    · 30′
  8. C08
    Benchmark Contamination and Decontamination
    LLM ArchitectureCapability EvaluationBenchmarks+4
    · 30′
  9. C09
    Benchmark Score Redundancy
    LLM ArchitectureCapability EvaluationBenchmarks+3
    · 50′
  10. C10
    Measuring Beyond Accuracy Saturation
    LLM ArchitectureCapability EvaluationBenchmarks+3
    · 65′
  11. C11
    Compute-Controlled Benchmarking
    LLM ArchitectureCapability EvaluationBenchmarks+2
    · 87′
  12. C12
    Large-Scale Test-Time Compute
    LLM ArchitectureTest Time ComputeInference Scaling+2
    · 52′
  13. C13
    The Open-Weight Frontier Gap
    LLM ArchitectureOpen WeightsCapability Evaluation+1
    · 55′
  14. C14
    DRACO Benchmark
    BenchmarksCapability EvaluationDeep Research+1
    · 13′
  15. C15
    AI Accelerating AI Development
    GovernanceAI RdProductivity+2
    · 33′
  16. C16
    AI R&D Autonomy Evaluation (AECI)
    GovernanceAI RdCapability Evaluation+2
    · 43′
  17. C17
    Task Time-Horizon Scaling
    LLM ArchitectureCapability EvaluationBenchmarks+1
    · 30′