Repository navigation
[Eval 2/2] evaluate next-frame open-loop chunks and videos from measured anchors - #299
Draft
AnikethCheluva wants to merge 1 commit into
Conversation
This was referenced Oct 23, 2026
Author
This was referenced Oct 23, 2026
This was referenced Oct 23, 2026
AnikethCheluva
force-pushed
the
codex/arc-rebuild/19-distance-budget-dtw
branch
from
October 23, 2026 20:06
747c2dd to
ba5a2ba
Compare
AnikethCheluva
force-pushed
the
codex/arc-rebuild/20-open-loop-stream-evaluation
branch
from
October 23, 2026 20:06
361c13e to
5fd5e81
Compare
AnikethCheluva
force-pushed
the
codex/arc-rebuild/20-open-loop-stream-evaluation
branch
from
October 23, 2026 20:38
5fd5e81 to
c571874
Compare
This was referenced Oct 23, 2026
Draft
AnikethCheluva
force-pushed
the
codex/arc-rebuild/19-distance-budget-dtw
branch
from
October 23, 2026 22:02
00fbd9c to
94b75aa
Compare
AnikethCheluva
force-pushed
the
codex/arc-rebuild/20-open-loop-stream-evaluation
branch
from
October 23, 2026 22:02
c571874 to
56df89f
Compare
This was referenced Oct 26, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.

Stack context
Eval 2/2 of the Cartesian ARC arrival-v2 review. Main stack: #311 (38 levels). Optional projection: #317, outside the main tip's ancestry. Parent:
codex/arc-rebuild/19-distance-budget-dtw.What and why
Thread each source sample's measured ARC anchor into decoding, metric records and videos. Compare t+1 onward for ARC, leave baseline same-index behavior intact. Retire chunking selectors; keep caller episode bounds.
Needed / not carried forward
This level supplies the capability named above. The integrated stack has one canonical Cartesian codec, fixed independent-stream execution and model-declared decode context; it does not carry duplicate E1/legacy codecs or selectable race/joint-distance algorithms. Independent planar/simulation codecs remain outside this cleanup. No tests, fixtures or local verification scripts are committed. No training job or hardware session is launched.
Compatibility and verification
The full integration intentionally changes ARC targets: source starts at the next frame, row zero is incoming travel, and timing is measured from current proprio. Fresh compatible normalization and new experiment identities are required. Old checkpoints retain their recorded sources; same tensor width does not make them compatible. Complete experiment menus activate at the config integration level; intermediate helper PRs are not production launch candidates.
External CPU checks: 114 codec checks; 457 staged-codec checks; 93 inference/evaluator context checks; 282 open-loop oracle checks; 27 diagnostic checks; numerical post-tokenization statistics and next-frame source/preflight checks. All 325 current YAMLs resolve; 24 ARC and four baseline config-to-decoder context constructions pass. The optional projection extension passed 14 dedicated checks. These are integration receipts, not claims that every intermediate PR ran a complete training or hardware test.