Yifan Yang 77ac33e8bf feat(sleep): multi-rollout contrastive reflection + token/time budget ...

The "脑补推演" core the user described — re-run the same task many times and
learn from the contrast between good and bad rollouts:

  - rollout.py: multi_rollout(task, k) runs K scored attempts; RolloutSet exposes
    best/worst/spread/pass_rate. contrastive_reflect picks the highest-spread
    tasks (some attempts passed, some failed — most informative) and asks the
    optimizer what the GOOD attempts did that the BAD ones didn't, distilling a
    general rule. Far stronger signal than a single failure.
  - consolidate(rollouts_k>1) uses contrastive reflection (falls back to
    single-shot reflect if it yields nothing).
  - budget.py: Budget(max_tokens|max_minutes) tracks spend; plan_depth() derives
    (nights, rollouts_k) from a token budget. run_gbrain gains --rollouts-k,
    --budget-tokens, --budget-minutes (auto-plans depth).

3 new tests (rollout stats, budget+plan, contrastive stub). 26 tests pass.

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

2026-06-08 14:31:51 +00:00

..

__init__.py

test: add unit test suite for core utility modules

2026-06-01 02:04:22 +08:00

test_json_utils.py

test: add unit test suite for core utility modules

2026-06-01 02:04:22 +08:00

test_qwen_backend.py

fix(model): forward Qwen timeout and only set enable_thinking when true

2026-06-07 07:41:35 -07:00

test_scoring.py

test: add unit test suite for core utility modules

2026-06-01 02:04:22 +08:00

test_sleep_engine.py

feat(sleep): multi-rollout contrastive reflection + token/time budget

2026-06-08 14:31:51 +00:00

test_types.py

test: add unit test suite for core utility modules

2026-06-01 02:04:22 +08:00