Repository navigation
World stage b2: comparisons, numeric fluents, numeric conditions - #27
Merged
Merged
Conversation
… the cleanup's follow-ups The world's comparisons of scalars (GREATER, ABOVE, BELOW over entities and category means, with ENTITY as the comparison class of every entity, derived/comparison_classes.csv, and the functions beside derive in the runtime); numeric fluents (NUMFL.<n> symbols with a range, an initial distribution, and a drift; numeric effects, numeric threshold literals in preconditions and in conditions, derived Boolean fluents over numeric thresholds, comparisons at a time point with RANGE, INCREASE, and DECREASE; the order clamp(clamp(v + e) + d) within a step; histories that record every numeric change with its cause; scene.initial: redraw for numeric fluents); the numeric condition literals deferred from stage b1; the brute-force evaluator, ten new hand fixtures, the regenerated tiny fixtures, and the fixtures' README for the Rust runtime; the statistics; and the corpus kept correct without a word for any of it. Every numeric draw comes from a stream of its own (world:numeric, numeric_initial, numeric_effects, numeric_preconditions, numeric_conditions), by replacement of a Boolean draw at the share, so that numeric_fluents.count: 0 gives the stage b1 worlds byte for byte and the shares at 0 give their Boolean dynamics. The engineering choices are WM.E191 to WM.E206 in docs/DECISIONS.md, from docs/proposals/2026-10-12-world-stage-b2-decisions.md, which also records the changed expectations, the spec edits, the two-place shares, the run times, the corpus's figures, the open questions, and the notes for stage b3. The four follow-ups of the process cleanup (Jon's rulings 3 to 6, recorded in that proposal; ruling 3 is PR.2): the random-streams rule of CLAUDE.md, the nec_none margin's comment, the world's interference tests back in the fast tier with pytest -n auto as its command, and the torch example's thread count. The default and tiny worlds and corpora are regenerated (the old folders moved aside as _pre_b2); the snapshot and the guides' figures are rewritten: every figure of the dynamics, the episodes, and the corpora moved with the new worlds (319 figures), and the baselines of every off setting with them. Run times: the default world 6.7 s (7.6 s in b1), the default corpus 157.6 s (157.4 s). The two-place share on seeds 1 to 5: 0.347, 0.333, 0.440, 0.072, 0.501 (b1: 0.313, 0.519, 0.379, 0.119, 0.509). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
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
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.
Stage b2 of the world-and-language refactor, with the process clean-up's follow-ups. See docs/proposals/ for the b2 proposal file.