GuardianAILAB1 — Deterministic ContractSCC / REP
Baseline suite isolating structural, boundary, and deterministic state-transition failures under recursive LLM interaction.
Run OFF - GuardianAI IDLE
Website
Comparator: Raw Byte Equality (No Normalization)Model: openai/gpt-oss-20b
0.00
0.00

Passive Mode: no retries. The first model output is the final result.

Execution: Passive (Default)

Comparator: Raw Byte-Exact

Instrument: Contract Enforcement

PAUSE handling: no retry path in passive mode.

Script Setup

Script file: public/scripts/online_hellaswag_label_contract.jsonlView raw on GitHub

GuardianAI Deterministic Contract Compliance Test

Contract compliance lab (not a reasoning benchmark)

Raw-byte strict mode

Panel 1 - Contract Definition

Expected output (strict contract field)

[expected literal not available yet]
Prompt-declared literalN/A
Comparison modeRaw Byte-Exact

Panel 2 - Live Output

Length0 chars
Byte count0 bytes

Escaped output literal

[no output yet]
UTF-8 bytes[]

Panel 3 - Contract Result

Initial exact matchN/A
Exact MatchN/A
Prompt-aligned matchN/A
Semantic classN/A
Hard failures0/0
GateN/A
Retries used0

Panel 4 - Statistical Summary

After N trials0
Raw failure rate (baseline)N/A
Hard semantic rate (baseline)N/A
Format-only mismatch rateN/A
95% CI (raw-byte)n/a

LLM BASELINE

No conversation yet. Press Run Experiment.

Deterministic Contract Monitor

Contract Monitor Active

MODE: PASSIVETURNS: 0PAUSES: 0RETRIES: 0CONSTRAINTS: 0OBSERVATIONS: 0

No turn telemetry yet.