docs: reframe framework_advantages opening + correct the knee-point selector across docs - #136
Merged
Merged
Conversation
…lector framework_advantages.md framed the reader's choice as 'our wrapping layer vs. calling dspy.GEPA directly'; the real choice is adopting this framework vs. running stock DSPy + GEPA as released. Reword the opening accordingly and drop the loose 'fork'/'wrapping layer' language. Make the paired-bootstrap deploy gate the headline advantage (matching the README) and rewrite the knee-point section honestly: the e-band selector is a measured no-op on the default val-best path (it defers to GEPA's argmax; the e-band survives only behind --knee-point-strategy smallest), per the deploy-gate calibration. Framed as a credibility signal rather than a buried caveat. Correct the stale 'parsimony pick is the default' framing in architecture.md, workflows.md, and components.md — including the tool pipeline, whose diagram and step list showed a select_knee_point call the tool path never makes (it defers to val-argmax). Reframe the analysis doc's 'Superseded' banner as honest design context.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
framework_advantages.mdopening: the reader's decision is adopt this framework vs. run stock DSPy + GEPA as released, not "our wrapping layer vs. callingdspy.GEPAdirectly." Dropped the loose "fork"/"wrapping layer" wording in the closing section too.Why
The ε-band knee-point selector was calibrated to a no-op on the default
val-bestpath — 10/10 ε-mode agreement on a regenerated corpus; the band-walker re-picks GEPA's argmax every time. The code already defers to GEPA's argmax; the docs lagged, andframework_advantages.mdeven headlined the no-op as the framework's #1 advantage. It now reads as a calibrated-out feature that survives only behind--knee-point-strategy smallest— a rigor signal, not a buried caveat.Changed
framework_advantages.md— opening reframe, gate as the headline advantage, honest knee-point section, fixed analysis-doc links and closing wording.architecture.md,workflows.md,components.md— corrected the stale "parsimony pick is the default" framing, including the tool pipeline, whose diagram and step list showed aselect_knee_pointcall the tool path never makes (it defers to val-argmax).research/knee_point_analysis.md— "Superseded" banner reframed as honest design context, preserving the literature recon.Docs-only; no code touched.