The Five Receiver Primitives
Every RPCS-1 recommendation is computed from five receiver primitives. Each primitive is a scalar in [0, 100], derived deterministically from your environment inputs. Together they characterise the receiver profile that is structurally stable in your agent's environment.
Temporal Integration
[0, 100]
Driven by
Environmental entropy (Matching Principle: TI ~ 1/H)
Maps to
max_tokens, context_strategy
Low value
Short attention window — agent processes recent signals only (frequent_grounding strategy)
High value
Long attention window — agent integrates deep history (long_window strategy)
Example
Chaotic environment (H=0.95) → TI ≈ 10. Stable environment (H=0.2) → TI ≈ 77.
Signal Gain
[0, 100]
Driven by
Stakes (primary), environmental predictability (adjustment)
Maps to
temperature (inversely), top_p
Low value
Low amplification — agent is conservative, less responsive to weak signals. Maps to LOWER temperature.
High value
High amplification — agent responds strongly to all signals. Maps to HIGHER temperature.
Example
catastrophic stakes → SG ≈ 20 → temperature ≈ 0.8 (more cautious sampling). low stakes → SG ≈ 75 → temperature ≈ 0.25.
Filtering Threshold
[0, 100]
Driven by
Stakes (primary), commitment style (adjustment)
Maps to
tool_use_strategy, system_prompt_additions
Low value
Low gating — agent acts readily on incoming signals (aggressive tool use).
High value
High gating — agent verifies before acting (explicit_confirmation tool strategy).
Example
catastrophic stakes + cautious style → FT = 95 → explicit_confirmation + high_stakes system prompt.
Update Elasticity
[0, 100]
Driven by
Environmental entropy (primary), context relevance (adjustment)
Maps to
retry_strategy
Low value
Low elasticity — agent resists revising its model (minimal retry).
High value
High elasticity — agent updates readily on new information (aggressive retry).
Example
chaotic entropy + short context → UE = 92 → aggressive retry. stable + long context → UE ≈ 17 → minimal retry.
Ambiguity Resolution
[0, 100]
Driven by
Commitment style (primary), stakes (adjustment)
Maps to
tool_use_strategy (combined with FT)
Low value
Low resolution — agent defers when uncertain, asks for clarification (ambiguity_caution prompt).
High value
High resolution — agent commits under uncertainty, resolves ambiguity aggressively.
Example
cautious + catastrophic → AR = 10. decisive + low → AR = 75.