From: Jérôme Benoit Date: Mon, 6 Oct 2025 13:01:07 +0000 (+0200) Subject: docs(reward): dedupe help text and align exit_piecewise_grace description X-Git-Url: https://git.piment-noir.org/?a=commitdiff_plain;h=c06795f1f1cb1cbb05089eb3809ad5465ea6b3c9;p=freqai-strategies.git docs(reward): dedupe help text and align exit_piecewise_grace description --- diff --git a/ReforceXY/reward_space_analysis/reward_space_analysis.py b/ReforceXY/reward_space_analysis/reward_space_analysis.py index 6456db6..ae39f18 100644 --- a/ReforceXY/reward_space_analysis/reward_space_analysis.py +++ b/ReforceXY/reward_space_analysis/reward_space_analysis.py @@ -163,13 +163,14 @@ DEFAULT_MODEL_REWARD_PARAMETERS_HELP: Dict[str, str] = { "holding_penalty_power": "Power applied to holding penalty scaling.", "exit_factor_mode": "Time attenuation mode for exit factor.", "exit_linear_slope": "Slope for linear exit attenuation.", - "exit_piecewise_grace": "Grace region for piecewise exit attenuation.", - "exit_piecewise_slope": "Slope after grace for piecewise mode.", + # exit_piecewise_grace: duration ratio boundary; >1 extends full-strength region + "exit_piecewise_grace": "Grace boundary (duration ratio; >1 extends no-attenuation region).", + "exit_piecewise_slope": "Slope after grace for piecewise mode (0 = flat).", "exit_power_tau": "Tau in (0,1] to derive alpha for power mode.", "exit_half_life": "Half-life for exponential decay exit mode.", "efficiency_weight": "Weight for efficiency factor in exit reward.", "efficiency_center": "Center for efficiency factor sigmoid.", - "win_reward_factor": "Amplification for pnl above target.", + "win_reward_factor": "Amplification for pnl above target (no hard cap; asymptotic).", "pnl_factor_beta": "Sensitivity of amplification around target.", "check_invariants": "Boolean flag (true/false) to enable runtime invariant & safety checks.", "exit_factor_threshold": "If |exit factor| exceeds this threshold, emit warning.",