docs: weekly sync 08-20 deck; bridge demo-free single-seed configs - #151
Merged
Conversation
yichao-liang
force-pushed
the
sync-slides-0820
branch
from
August 20, 2026 21:04
3a71f3b to
95435b3
Compare
yichao-liang
force-pushed
the
bridge-learning-integrity
branch
from
August 20, 2026 21:04
d6736cb to
f5738c1
Compare
yichao-liang
force-pushed
the
sync-slides-0820
branch
from
August 20, 2026 21:22
95435b3 to
12ef1c4
Compare
yichao-liang
force-pushed
the
bridge-learning-integrity
branch
from
August 20, 2026 21:22
f5738c1 to
619cd5a
Compare
yichao-liang
force-pushed
the
sync-slides-0820
branch
from
August 20, 2026 21:51
12ef1c4 to
b85b55c
Compare
The agent arm's demonstrator plans without the simulator (bilevel_plan_without_sim) and test tasks get a single solve attempt (agent_solve_max_attempts: 1).
Per-env demo clips (domino/fan test episodes, a bridge oracle_hybrid_sim solve, a freshly re-recorded boil oracle demo - the original learned-run videos predate this repo's logs) plus two bridge 1x3 seed mosaics built by the new make_bridge_mosaics.py. All clips are faststart-remuxed so they stream when the deck is served over HTTP.
The week in one deck: the novelty claim, where the four domains stand (with per-env demo videos), the agent's three-phase design, and the bridge story - oracle_hybrid_sim 12/12, what the first learning runs exposed (ground-truth leaks, privileged goal atoms, unbounded Waits), the fixes, and the honest relaunch. Every slide is annotated with the week's commits. make_standalone.py rebuilds the shareable single-file version (reveal.js, fonts, and all assets inlined as data URIs); the generated *_standalone.html is a build artifact and stays uncommitted.
max_initial_demos 1 -> 0 for bridge (the AL arm learns from its own exploration only, matching the other agent domains) and NUM_SEEDS 3 -> 1 in common.yaml for the current single-seed launch cadence.
'submmited' -> 'submitted', 'helps it's' -> 'help its', a missing colon after 'never terminated', and the validation gate escalates after a flaky REJECTION, not a flaky pass.
yichao-liang
force-pushed
the
sync-slides-0820
branch
from
August 20, 2026 22:20
b85b55c to
3920b2b
Compare
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.
Part 6/6 of the bridge-domain stack (based on #150).
Review fixes folded in: deck typos and the flaky-gate wording (escalation follows a flaky rejection, not a flaky pass).
🤖 Generated with Claude Code