Order-2 ancestral denoised extrapolation (automatic, zero extra cost):
The early Euler-A phase now reuses the previous step's denoised output
to extrapolate a better score direction estimate before each ancestral
step, equivalent to DPM-Solver++ 2M applied to the stochastic phase.
The state is reset after any motion pulse since a pulse modifies x and
invalidates the previous estimate.
Progressive correction_blend in quality phase (automatic, speeds up):
The first half of quality-phase steps now use correction_blend=0
(single-eval Euler ODE); the second half use the configured blend
(full 2S correction). _hybrid_ode_step already returns early at blend=0
with no second model call. For a 30-step run with 35% quality phase this
saves roughly 4-5 model calls (~12% faster) while keeping the expensive
2S correction where sigma is lowest and it has the most impact.
Adaptive eta decay (new eta_final parameter, default 1.0):
When eta_final is set below eta, ancestral noise strength decays linearly
toward eta_final as sigma approaches the quality boundary, based on
sigma proximity (quality_sigma_start / sigma). This reduces stochastic
variation right before the deterministic quality phase, giving a cleaner
hand-off. Default 1.0 preserves existing behaviour.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Fix _v2_find_perfect_example_for_intent: was reading non-existent
loved_delta_sources field; now correctly reads perfect_anchors/base
and loved_variants using the actual positive_prompt/encoded_prompt keys
- Fix _v2_build_v2_recommendations: removed global phrase_memory liked
phrases that would inject unrelated-intent vocabulary for universal keys;
recommendations now only include axis feedback and intent-filtered
repair candidates
- Fix _v2_update_streaks in Prompt only mode: avg_reward_ema and
good/bad streaks are now skipped in Prompt only mode so they cannot
contaminate conditioning adaptation strength in subsequent Refine runs;
last_rating_label and last_missing_axes still update for repair continuity
- Fix advisor rating label on reset/first run: advisor now receives
"No previous output (first run or session reset)" instead of the
user-set rating widget value when has_previous_run is False, preventing
the advisor from treating a stale rating as evidence
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Remove hardcoded Gemma chat tokens; advisor prompts now use plain
structured text compatible with any CLIP text generator
- Raise advisor generation limit from 384 to 800 tokens (internal)
- Repair candidates are now passed as suggestions to the advisor before
being applied; rule-based repair applies only if the advisor is off or
produces no valid output
- Add optional feedback_prompt input for user-provided natural-language
feedback, injected at highest priority in the advisor system prompt
- Add Prompt only execution mode: full prompt shaping with conditioning
vectors passed through unchanged
- Rename advisor modes to Off / Only diagnostics / Only prompt / Full;
Only prompt suppresses the diagnostic line for clean output
- Advisor detects when its own previous output caused a bad rating and
notes this explicitly in the next system prompt
- Inject perfect-rated prompt example into advisor system prompt when
one exists for the current intent family
- Inject V2 phrase-memory recommendations into advisor system prompt;
user feedback takes priority over these
- Guard against stale rating attribution when user intent changes between
runs (body similarity < 0.50 skips repair application)
- Advisor no longer skips on Perfect rating; runs in analysis-only mode
to capture what worked
- Previous advisor diagnostic is fed back into the next run's system
prompt, closing the learning loop
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
