H
Howardism
Howardism · Vol. 03Plate I · No. 01

Evaluation, tagged.

Notes52TagEvaluationOldest8 May 2026Newest1 Oct 2026

Every article tagged evaluation, newest first.

  1. C01
    Verbalized-Confidence Soft Scoring for LLM Judges
    LLM As A JudgeCalibrationVerbalized Confidence+2
    · 16′
  2. C02
    Version-Dependent Judge Error
    EvaluationLLM As A JudgeMeasurement+3
    · 18′
  3. C03
    Typed Decision Verifiers
    EvaluationVerificationLLM As A Judge+1
    · 20′
  4. C04
    Long-Horizon Agent Failure Signature
    Agent EngineeringFailure ModesEvaluation+2
    · 16′
  5. C05
    Sandbagging Elicitation (Reference & Context Grafting)
    InterpretabilityAlignmentSafety+2
    · 13′
  6. C06
    Embedded Evaluation
    AlignmentEvaluationGovernance+2
    · 15′
  7. C07
    Cross-Model Error Entanglement
    EvaluationLLM As A JudgeReliability+2
    · 55′
  8. C08
    Evaluation Horizon Versus Release Cadence
    EvaluationSafetyLong Horizon+2
    · 18′
  9. C09
    The OpenAI / Hugging Face Intrusion (July 2026)
    SecurityIncident ResponseThreat Landscape+2
    · 68′
  10. C10
    Unsanctioned Agent Message Boards
    SecurityMulti AgentCovert Channel+3
    · 50′
  11. C11
    Skill Lift
    EvaluationAgent SkillsAblation+3
    · 25′
  12. C12
    AI-Assisted Error Analysis
    EvalsEvaluationError Analysis+3
    · 16′
  13. C13
    Agent Epistemic Vigilance
    AlignmentSafetyMulti Agent+2
    · 16′
  14. C14
    Multiagent Turf War
    AlignmentSafetyMulti Agent+2
    · 17′
  15. C15
    GDPval Benchmark
    EvaluationBenchmarksEconomic Impact+1
    · 39′
  16. S16
    Guarantees That Degrade at Deployment: Action-Space Soundness, Admissibility Without Effect, and a Vendor-Coupled Security Framework
    DerivedAgent EngineeringSecurity+3
    · 34′
  17. C17
    Weak-Verifier Ensembling
    EvaluationLLM As A JudgeTest Time Compute+2
    · 51′
  18. C18
    Deterministic Engineering for Agent Code Review
    Agent EngineeringCode ReviewHarness+3
    · 44′
  19. C19
    Cheating in Capability Evaluations
    AlignmentSafetyEvaluation+3
    · 60′
  20. C20
    Misalignment in Production Agent Traffic
    AlignmentSafetyAgents+3
    · 45′
  21. C21
    Same-Model Review Blindness
    Code ReviewAI Coding WorkflowEvaluation+1
    · 53′
  22. C22
    Self-Negotiated Contracts Between Agents
    AlignmentSafetyMulti Agent+3
    · 28′
  23. C23
    Task Gaming
    AlignmentSafetyReward Hacking+2
    · 52′
  24. C24
    User Awareness
    AlignmentSafetyEvaluation+2
    · 35′
  25. C25
    Unsanctioned Action in Capability Evaluations
    SafetyAlignmentEvaluation+3
    · 73′
  26. C26
    Automated Failure Attribution
    Agent EngineeringFailure ModesMulti Agent+2
    · 48′
  27. C27
    Harness-Induced Belief Divergence
    Agent EngineeringHarnessEvaluation+2
    · 21′
  28. C28
    Stopping Under a Noisy Verifier
    Agent EngineeringEvaluationLoops+2
    · 40′
  29. C29
    Agent-Authored Harness Optimization
    Agent EngineeringHarnessEvaluation+2
    · 127′
  30. C30
    Deterministic Pre-Execution Gates
    Agent EngineeringHarnessFailure Modes+2
    · 39′
  31. C31
    Layerwise Omission Attribution
    Agent EngineeringFailure ModesContext Management+2
    · 26′
  32. C32
    Autonomous Intrusion
    SecurityIncident ResponseThreat Landscape+3
    · 52′
  33. C33
    Model Organisms
    InterpretabilityAlignmentSafety+2
    · 14′
  34. C34
    AI-to-AI Coercion
    AlignmentSafetyEvaluation+3
    · 20′
  35. C35
    Motivated Mislabeling
    AlignmentSafetyEvaluation+2
    · 7′
  36. C36
    Promise-Breaking in Multi-Agent Games
    AlignmentSafetyMulti Agent+3
    · 17′
  37. C37
    Confident But Unsure
    AlignmentHonestyCalibration+2
    · 24′
  38. C38
    Unproductive Self-Verification
    CapabilityTest Time ComputeAgents+2
    · 18′
  39. C39
    Usage-Telemetry Classifier Validation
    EvaluationMeasurementLLM As A Judge+3
    · 22′
  40. C40
    Reference-Free Judge Over-Crediting
    EvaluationLLM As A JudgeBenchmarks+3
    · 46′
  41. C41
    LLM-Judge Validation
    EvaluationLLM As A JudgeBenchmarks+2
    · 86′
  42. C42
    Agent Quality Flywheel
    EvaluationEvalsAgent Engineering+3
    · 25′
  43. C43
    Failures That Look Like Success
    EvaluationAgent EngineeringFailure Modes+1
    · 41′
  44. E44
    Gemini Enterprise Agent Platform
    EntityPlatformGoogle+1
    · 4′
  45. C45
    Optimizer–Evaluator Decoupling
    EvaluationAgent EngineeringReward Hacking+1
    · 58′
  46. C46
    Deployment Simulation
    AlignmentSafetyEvaluation+2
    · 29′
  47. C47
    LLM-as-a-Judge
    EvaluationLLM As A JudgeBenchmarks+1
    · 67′
  48. C48
    Production-Sourced Evaluation
    EvaluationBenchmarksData Provenance+1
    · 32′
  49. C49
    Agentic Honesty & Diligence
    AlignmentHonestyAgents+2
    · 35′
  50. C50
    Automated Behavioral Audit
    AlignmentSafetyEvaluation+2
    · 26′
  51. C51
    Evaluation Awareness & Grader Gaming
    AlignmentSafetyEvaluation+2
    · 65′
  52. C52
    Agentic Misalignment (AM)
    AlignmentSafetyEvaluation+2
    · 27′