research(ai): encoder knob-space Pareto frontiers — analysis scaffold (ADR-0305 / Research-0077) by lusoris · Pull Request #400 · lusoris/vmaf

lusoris · 2026-05-05T19:12:36Z

Companion to Research-0063 (CQ vs VBR stratification finding). This PR ships the analysis methodology + scripts; actual headline findings land via follow-up commit when the sweep completes (~3h ETA).

Summary

Ships the methodology + analysis scaffold for the 12,636-cell encoder knob sweep (9 sources × 6 codec families × 3 rate-control modes × ~78 knob combinations per codec) that drives tools/vmaf-tune/codec_adapters/* recipe defaults.
Decision (ADR-0305): stratify the dominance hull on (bitrate_kbps, vmaf_score) per (source, codec, rc_mode) slice — never as a single global frontier — because Research-0063 showed a global hull collapses the rate-control flip and produces consensus recipes that regress NVENC h264/hevc by ~4 VMAF at cq=30 against the bare encoder defaults. encode_time_ms is the tiebreaker on hull-boundary rows.
New ai/scripts/analyze_knob_sweep.py consumes comprehensive.jsonl, computes per-slice Pareto frontiers, emits per-slice CSV tables + a markdown summary, and carries a detect_recipe_regressions(...) check that gates ship-candidate adapter defaults.
The comprehensive.jsonl sweep file is generated locally and lives under runs/phase_a/full_grid/ (gitignored — not shipped in this PR). The headline-findings table in Research-0077 is currently scaffolded with TBD pending sweep completion; a follow-up commit on this branch flips it once the sweep finishes.

Six deep-dive deliverables (ADR-0108)

(1) Research digest: docs/research/0077-encoder-knob-space-pareto-frontiers.md
(2) Decision matrix: ADR-0305 §Alternatives considered (per-(source, codec, rc_mode) stratified Pareto [chosen] vs global Pareto vs codec-only stratified)
(3) AGENTS.md invariant note: ai/AGENTS.md — knob-sweep corpus invariant ("recipes that regress vs the bare encoder at matched bitrate within the same slice MUST NOT ship as adapter defaults")
(4) Reproducer / smoke-test command: pytest ai/tests/test_knob_sweep_analysis.py -v
(5) CHANGELOG fragment: changelog.d/added/encoder-knob-space-pareto-analysis.md
(6) Rebase note: docs/rebase-notes.md §0305

Test plan

python3 -c "import ast; ast.parse(open('ai/scripts/analyze_knob_sweep.py').read())" — clean
pytest ai/tests/test_knob_sweep_analysis.py -v — 3 passed (test_pareto_frontier_smoke, test_stratification_keys, test_recipe_regression_detection)
black --check ai/ — clean (120 files)
ruff check ai/ — clean
Sweep-completion follow-up commit: populate Research-0077 §Headline findings + open the recipe-default workstream PR for tools/vmaf-tune/codec_adapters/* (gated on detect_recipe_regressions returning empty for every shipped recipe)

Notes

Status: Draft until the sweep completes and the headline findings land.
ADR-0305 status: Proposed (will flip to Accepted in the same follow-up commit that adds the populated findings).
No Netflix golden assertions touched; no upstream-shared paths touched.

🤖 Generated with Claude Code

Copilot

Pull request overview

Adds a fork-local analysis scaffold to compute per-(source, codec, rc_mode) Pareto frontiers over a large encoder knob sweep, plus the accompanying ADR/research documentation and a regression gate to prevent shipping recipe defaults that lose to bare encoder settings at matched bitrate.

Changes:

Introduces ai/scripts/analyze_knob_sweep.py to load a JSONL sweep, stratify by (source, codec, rc_mode), compute Pareto hulls, and report CSV + markdown summaries, including a regression-detection gate.
Adds unit tests with a synthetic JSONL fixture to exercise stratification, hull selection (incl. encode-time tiebreak), and regression detection.
Documents the methodology/decision via ADR-0305 + Research-0077, and records the change in ai/AGENTS.md, rebase notes, and a changelog fragment.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
docs/research/0077-encoder-knob-space-pareto-frontiers.md	New research digest describing sweep axes, per-slice frontier methodology, and scaffolded findings.
docs/rebase-notes.md	Adds rebase ledger entry for ADR-0305 / sweep analysis scaffold.
docs/adr/README.md	Adds ADR-0305 row to ADR index table (but this file is generated from fragments).
docs/adr/0305-encoder-knob-space-pareto-analysis.md	New ADR documenting the stratified Pareto-frontier decision and consequences.
changelog.d/added/encoder-knob-space-pareto-analysis.md	Changelog fragment announcing the analysis scaffold and tests.
ai/tests/test_knob_sweep_analysis.py	New tests validating hull computation, stratification keys, and regression detection.
ai/scripts/analyze_knob_sweep.py	New analysis script: load JSONL, compute per-slice Pareto hulls, write CSV + summary, detect regressions vs bare defaults.
ai/AGENTS.md	Adds a “knob-sweep corpus invariant” tying adapter-default eligibility to per-slice hulls + non-regression gating.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

 | [ADR-0283](0283-vmaf-tune-videotoolbox-adapters.md) | Apple VideoToolbox codec adapters for `tools/vmaf-tune/`. Adds `H264VideoToolboxAdapter` + `HEVCVideoToolboxAdapter` (and a shared `_videotoolbox_common.py` for the `-q:v` 0..100 quality knob + nine-name preset → `-realtime` boolean mapping) along the same one-file-per-codec contract NVENC / AMF / QSV already use. AV1 hardware encoding intentionally omitted (unavailable on Apple Silicon as of 2026). Tests mock `subprocess.run` so Linux CI stays green; macOS end-to-end is left to contributors with VideoToolbox available locally. The originally-coupled 16-slot codec-vocab schema expansion is deferred to a follow-up PR awaiting a fresh `fr_regressor_v2` retrain (ship-gate per ADR-0235 + ADR-0291). Companion research digest [`docs/research/0074-vmaf-tune-videotoolbox-adapters.md`](../research/0074-vmaf-tune-videotoolbox-adapters.md). | Accepted | tooling, ai, ffmpeg, codec, hardware-encoder, apple, fork-local |
 | [ADR-0297](0297-vmaf-tune-sample-clip.md) | `vmaf-tune --sample-clip-seconds N` — opt-in sample-clip mode that encodes/scores only the centre N-second window of each source per grid cell instead of the full reference, scaling per-cell wall time roughly linearly with slice length (e.g. ~6x speedup at `N=10` against a 60-second source). FFmpeg input-side `-ss <start> -t <N>` cuts the rawvideo demuxer at the slice boundary; the libvmaf CLI's `--frame_skip_ref` / `--frame_cnt` mirror the same window on the score side so VMAF compares matching frames without slicing the reference YUV on disk. Centre-anchored placement (naive scaffold; TransNet V2-based smart placement is a follow-up). Each emitted row carries `clip_mode = "sample_<N>s"` or `"full"` so Phase B (target-VMAF bisect) and Phase C (per-title CRF predictor) can filter, weight, or epilogue-rescore. Corpus schema bumps additively to `SCHEMA_VERSION = 2`. Expected accuracy delta ~1–2 VMAF points on diverse content, ~0.3–0.5 on uniform content. Companion to [ADR-0237](0237-quality-aware-encode-automation.md) Phase A. | Accepted | tooling, ffmpeg, vmaf-tune, fork-local |
 | [ADR-0301](0301-vmaf-tune-sample-clip.md) | `vmaf-tune --sample-clip-seconds N` — opt-in sample-clip mode that encodes/scores only the centre N-second window of each source per grid cell instead of the full reference, scaling per-cell wall time roughly linearly with slice length (e.g. ~6x speedup at `N=10` against a 60-second source). FFmpeg input-side `-ss <start> -t <N>` cuts the rawvideo demuxer at the slice boundary; the libvmaf CLI's `--frame_skip_ref` / `--frame_cnt` mirror the same window on the score side so VMAF compares matching frames without slicing the reference YUV on disk. Centre-anchored placement (naive scaffold; TransNet V2-based smart placement is a follow-up). Each emitted row carries `clip_mode = "sample_<N>s"` or `"full"` so Phase B (target-VMAF bisect) and Phase C (per-title CRF predictor) can filter, weight, or epilogue-rescore. Corpus schema bumps additively to `SCHEMA_VERSION = 2`. Expected accuracy delta ~1–2 VMAF points on diverse content, ~0.3–0.5 on uniform content. Companion to [ADR-0237](0237-quality-aware-encode-automation.md) Phase A. | Accepted | tooling, ffmpeg, vmaf-tune, fork-local |
+| [ADR-0305](0305-encoder-knob-space-pareto-analysis.md) | Encoder knob-space Pareto-frontier analysis scaffold. Companion to [Research-0063](../research/0063-encoder-knob-space-cq-vs-vbr-stratification.md) (CQ vs VBR stratification finding). Decision: stratify the dominance hull on `(bitrate_kbps, vmaf_score)` **per `(source, codec, rc_mode)` slice** — never as a single global frontier — because Research-0063 showed a global hull collapses the rate-control flip and produces consensus recipes that regress NVENC h264/hevc by ~4 VMAF at cq=30 against bare encoder defaults. `encode_time_ms` is the tiebreaker on hull-boundary rows. The 12,636-cell sweep (9 sources × 6 codec families × 3 rc_modes × ~78 knob combinations per codec) is generated locally and lives at `runs/phase_a/full_grid/comprehensive.jsonl` (gitignored). New `ai/scripts/analyze_knob_sweep.py` computes per-slice frontiers + emits per-slice CSVs and a markdown summary, plus a regression-detection check (`detect_recipe_regressions`) that gates ship-candidate recipes — recipes that lose VMAF vs the bare encoder at matched bitrate within the same slice MUST NOT ship as `tools/vmaf-tune/codec_adapters/*` defaults. New `ai/AGENTS.md` knob-sweep corpus invariant pins this. Companion research digest: [`docs/research/0077-encoder-knob-space-pareto-frontiers.md`](../research/0077-encoder-knob-space-pareto-frontiers.md). Headline findings populate the Research-0077 §Headline-findings table via a follow-up commit when the sweep completes (~3h ETA). | Proposed | ai, vmaf-tune, research, encoder, pareto, fork-local |


@@ -0,0 +1,129 @@
+# Research-0077 — Encoder knob-space Pareto frontiers (per source × codec × rc_mode)
+
+- **Status**: Adopted by [ADR-0305](../adr/0305-encoder-knob-space-pareto-analysis.md)


+   listing the hull rows + the dominated rows directly above the
+   hull at each bitrate.


+# for hull-membership purposes. Calibrated against the per-frame VMAF
+# noise floor (~0.1 points) and bitrate quantisation in libavformat
+# muxers (~1 kbps).


+    """One cell of the knob sweep.
+
+    The schema mirrors `tools/vmaf-tune/` Phase A SCHEMA_VERSION = 2
+    (ADR-0301 sample-clip mode + ADR-0297 multi-codec dispatcher), so a


+
+- **Status**: Proposed
+- **Date**: 2026-05-05
+- **Deciders**: lusoris, Claude


…(ADR-0313) (#410) * ci(policy): Required Checks Aggregator — unblock doc/Python-only PRs (ADR-0313) The 23-named-required-check posture (ADR-0037) deadlocks doc/Python-only PRs: the C-build matrix path-filter-skips on their diffs, but branch protection counts a path-filter-skip + a never-ran-at-all as not satisfying the required-check. PR #400 hit this concretely (10/23 succeeded; 13/23 either skipped or never reported; gh pr merge returned "the base branch policy prohibits the merge"). Aggregator is one workflow with no path filter. It polls up to 8 minutes for sibling workflows to register, then verifies each named check on the head SHA reported success/skipped/neutral (or didn't appear at all, which is the documented path-filter rejection semantics). Aggregator becomes the single branch-protection required check; the 23 individual workflows continue to run unchanged. Manual operator step at adoption (after this PR merges): gh api -X PUT "repos/lusoris/vmaf/branches/master/protection/required_status_checks" \ -F 'strict=true' -F 'contexts=["Required Checks Aggregator"]' Unblocks #400, #403, #404, #405, #406, #407 currently stuck on the deadlock. Per user popup direction 2026-05-05. Files: .github/workflows/required-aggregator.yml (new), docs/adr/0313-*.md (new), changelog.d/added/*.md (new), docs/adr/README.md (+1 row), docs/adr/_index_fragments/_order.txt (+1 line + new fragment), docs/rebase-notes.md §0313. * ci: retrigger after PR body cleanup * ci: retrigger after deliverables opt-out polarity fix --------- Co-authored-by: Lusoris <lusoris@pm.me>

…#413) First post-#410 run on PR #400 hit the 8-minute poll deadline while the Ubuntu gcc/clang DNN + Windows MSVC builds were still IN_PROGRESS and were declared FAILURE by the aggregator (their actual conclusion would have been success a few minutes later). The fork's slowest required builds take 12+ minutes; 30-minute poll deadline gives ~2x headroom. Job-level timeout-minutes raised to 35 so the GitHub-side kill doesn't pre-empt the poll. No semantic change to the aggregator logic; same accept-set (success/skipped/neutral, plus "never appeared" = path-filter rejection). Single-line numeric tweak. Co-authored-by: Lusoris <lusoris@pm.me>

Companion to Research-0063 (CQ vs VBR stratification finding, already merged). Ships the methodology + analysis scripts for the 12,636-cell encoder knob sweep (9 sources x 6 codec families x 3 rate-control modes x ~78 knob combinations per codec) that drives tools/vmaf-tune/codec_adapters/* recipe defaults. Headline findings on the populated Pareto frontiers land via a follow-up commit when the sweep completes (~3h ETA). Decision (ADR-0305): stratify the dominance hull on (bitrate_kbps, vmaf_score) per (source, codec, rc_mode) slice -- never as a single global frontier -- because Research-0063 showed a global hull collapses the rate-control flip and produces consensus recipes that regress NVENC h264/hevc by ~4 VMAF at cq=30 against the bare encoder defaults. encode_time_ms is the tiebreaker on hull-boundary rows. Files: - docs/adr/0305-encoder-knob-space-pareto-analysis.md (Proposed) - docs/research/0077-encoder-knob-space-pareto-frontiers.md - ai/scripts/analyze_knob_sweep.py (per-slice Pareto hull + regression-detection check; emits per-slice CSVs + summary.md) - ai/tests/test_knob_sweep_analysis.py (synthetic 20-row JSONL fixture; 3 passing tests) - ai/AGENTS.md -- append knob-sweep corpus invariant - changelog.d/added/encoder-knob-space-pareto-analysis.md - docs/rebase-notes.md section 0305 The comprehensive.jsonl sweep file is generated locally and lives at runs/phase_a/full_grid/ (gitignored -- not shipped in this PR). Verification: - ast.parse on analyze_knob_sweep.py: clean - pytest ai/tests/test_knob_sweep_analysis.py -v: 3 passed - black ai/: clean (120 files) - ruff check ai/: clean Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…earch-0080) Runs the Research-0077 / ADR-0305 analysis script (ai/scripts/analyze_knob_sweep.py, ships in PR #400) over the 12,636-cell Phase A sweep at runs/phase_a/full_grid/comprehensive.jsonl and records the populated Pareto-hull populations + recipe-regression count per codec in Research-0080. ADR-0308 commits the fork to a structural-vs-content-dependent threshold for revision policy. Headline findings: - 162 realised slices (every slice has a populated hull). - 1,915 recipe-vs-bare regressions at default tolerances (bitrate_tol_pct=5, vmaf_tol=0.1). - CQP regression rate 6.6 % vs CBR 20.2 % / VBR 18.7 % re-confirms Research-0063 with hard numbers. - Top-15 aggregated bad-recipe cells all reproduce on all 9 corpus sources, clustered around h264_nvenc + bf3 / spatial_aq / full_hq under CBR/VBR plus a smaller hevc_nvenc + spatial_aq cluster. Decision (ADR-0308): a recipe regression is structural iff it reproduces on >=7 of 9 corpus sources within one (codec, rc_mode, recipe, preset, q) cell. Structural regressions are forbidden as tools/vmaf-tune/codec_adapters/* defaults and forbidden as vmaf-tune recommend outputs without explicit override; content-dependent regressions (1-6 sources) are filtered at recommend-time only via the per-slice hull lookup. The detector remains an offline gate (3-hour sweep too expensive for CI); promotion to a CI gate is deferred until a smaller stratified sample reproduces the structural patterns. Per-codec adapter revisions land as separate follow-up PRs for clean bisect signals. Six deep-dive deliverables (CLAUDE §11 / ADR-0108): - Research digest: docs/research/0080-encoder-knob-sweep-findings.md. - Decision matrix: ADR-0308 §Alternatives considered (4 options). - AGENTS.md invariant note: ai/AGENTS.md §Knob-sweep recipe-regression policy (cites ADR-0305 + ADR-0308). - Reproducer: pytest ai/tests/test_knob_sweep_analysis.py -v (script logic, ships in PR #400) + offline analyser run command in docs/rebase-notes.md §0308. - CHANGELOG fragment: changelog.d/changed/encoder-knob-sweep-findings.md. - Rebase note: docs/rebase-notes.md §0308. Constraints honoured: - Did not modify ai/scripts/analyze_knob_sweep.py (uses public API unchanged via a throw-away wrapper for the field-name rename). - Did not modify tools/vmaf-tune/codec_adapters/* (recipe revisions land in follow-up PRs). - Did not commit runs/ artefacts (.gitignore covers them). - Documentation-only; ~452 LOC against the 600-LOC budget. This PR queues behind PR #400 (ADR-0305 / Research-0077 / analysis script). Rebase target: master after PR #400 merges. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…(Research-0080) (#406) * docs(ai): encoder knob-sweep — Pareto hulls + recipe regressions (Research-0080) Runs the Research-0077 / ADR-0305 analysis script (ai/scripts/analyze_knob_sweep.py, ships in PR #400) over the 12,636-cell Phase A sweep at runs/phase_a/full_grid/comprehensive.jsonl and records the populated Pareto-hull populations + recipe-regression count per codec in Research-0080. ADR-0308 commits the fork to a structural-vs-content-dependent threshold for revision policy. Headline findings: - 162 realised slices (every slice has a populated hull). - 1,915 recipe-vs-bare regressions at default tolerances (bitrate_tol_pct=5, vmaf_tol=0.1). - CQP regression rate 6.6 % vs CBR 20.2 % / VBR 18.7 % re-confirms Research-0063 with hard numbers. - Top-15 aggregated bad-recipe cells all reproduce on all 9 corpus sources, clustered around h264_nvenc + bf3 / spatial_aq / full_hq under CBR/VBR plus a smaller hevc_nvenc + spatial_aq cluster. Decision (ADR-0308): a recipe regression is structural iff it reproduces on >=7 of 9 corpus sources within one (codec, rc_mode, recipe, preset, q) cell. Structural regressions are forbidden as tools/vmaf-tune/codec_adapters/* defaults and forbidden as vmaf-tune recommend outputs without explicit override; content-dependent regressions (1-6 sources) are filtered at recommend-time only via the per-slice hull lookup. The detector remains an offline gate (3-hour sweep too expensive for CI); promotion to a CI gate is deferred until a smaller stratified sample reproduces the structural patterns. Per-codec adapter revisions land as separate follow-up PRs for clean bisect signals. Six deep-dive deliverables (CLAUDE §11 / ADR-0108): - Research digest: docs/research/0080-encoder-knob-sweep-findings.md. - Decision matrix: ADR-0308 §Alternatives considered (4 options). - AGENTS.md invariant note: ai/AGENTS.md §Knob-sweep recipe-regression policy (cites ADR-0305 + ADR-0308). - Reproducer: pytest ai/tests/test_knob_sweep_analysis.py -v (script logic, ships in PR #400) + offline analyser run command in docs/rebase-notes.md §0308. - CHANGELOG fragment: changelog.d/changed/encoder-knob-sweep-findings.md. - Rebase note: docs/rebase-notes.md §0308. Constraints honoured: - Did not modify ai/scripts/analyze_knob_sweep.py (uses public API unchanged via a throw-away wrapper for the field-name rename). - Did not modify tools/vmaf-tune/codec_adapters/* (recipe revisions land in follow-up PRs). - Did not commit runs/ artefacts (.gitignore covers them). - Documentation-only; ~452 LOC against the 600-LOC budget. This PR queues behind PR #400 (ADR-0305 / Research-0077 / analysis script). Rebase target: master after PR #400 merges. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * ci: retrigger after deliverables N-prefix fix --------- Co-authored-by: Lusoris <lusoris@pm.me> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The 2026-05-06 merge train shipped 13 ADRs whose implementing PRs landed but Status was never bumped from Proposed to Accepted. Per docs/adr/README.md and ADR-0028, ADRs flip to Accepted once the deliverable lands. The train moved faster than the per-ADR Status edits could keep up; this PR catches up. Flipped: - ADR-0302 (#401, ENCODER_VOCAB v3 schema expansion) - ADR-0303 (#399, fr_regressor_v2 ensemble prod-flip gate) - ADR-0304 (#402, vmaf-tune fast-path Optuna TPE) - ADR-0305 (#400, knob-sweep Pareto analysis scaffold) - ADR-0307 (#404, vmaf-tune ladder default sampler) - ADR-0308 (#406, knob-sweep recipe-regression policy) - ADR-0309 (#405, ensemble retrain harness) - ADR-0311 (#408, libfuzzer harness expansion) - ADR-0313 (#410, CI Required Checks Aggregator) [table-format Status, sed-edited inline] - ADR-0314 (#412, vmaf-tune --score-backend=vulkan) - ADR-0316 (#414, cli_parse long-only-option assertion fix) - ADR-0317 (#415, CI Docker + FFmpeg-SYCL flake fix) - ADR-0319 (#422, ensemble LOSO trainer real impl) Already-Accepted (no change): ADR-0310 (#407), ADR-0312 (#425), ADR-0315 (skeleton, intentionally Proposed), ADR-0321 (#424).

lusoris force-pushed the research/encoder-knob-space-pareto-frontiers branch from b0f1a1d to 845d862 Compare May 5, 2026 20:07

lusoris marked this pull request as ready for review May 5, 2026 20:18

lusoris force-pushed the research/encoder-knob-space-pareto-frontiers branch from 845d862 to 126d4b7 Compare May 5, 2026 20:18

Copilot AI review requested due to automatic review settings May 5, 2026 20:18

Copilot started reviewing on behalf of lusoris May 5, 2026 20:24 View session

Copilot AI reviewed May 5, 2026

View reviewed changes

This was referenced May 5, 2026

research(ai): encoder knob-sweep — Pareto hulls + recipe regressions (Research-0080) #406

Merged

ci(policy): Required Checks Aggregator — unblock doc/Python-only PRs (ADR-0313) #410

Merged

lusoris force-pushed the research/encoder-knob-space-pareto-frontiers branch from 126d4b7 to d49028d Compare May 5, 2026 22:17

lusoris mentioned this pull request May 5, 2026

ci(aggregator): extend poll deadline 8m → 30m (post-bootstrap fix) #413

Merged

9 tasks

lusoris force-pushed the research/encoder-knob-space-pareto-frontiers branch from d49028d to 400a4d6 Compare May 5, 2026 22:51

ci: retrigger CI (MinGW64 transient + aggregator stale check_run)

5a8c890

lusoris merged commit 29818b2 into master May 5, 2026
55 checks passed

lusoris deleted the research/encoder-knob-space-pareto-frontiers branch May 5, 2026 23:40

lusoris mentioned this pull request May 6, 2026

docs(adr): bulk flip Proposed → Accepted for 13 merge-train ADRs #426

Merged

9 tasks

lusoris mentioned this pull request May 6, 2026

feat/vmaf tune score backend vulkan #436

Merged

9 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

research(ai): encoder knob-space Pareto frontiers — analysis scaffold (ADR-0305 / Research-0077)#400

research(ai): encoder knob-space Pareto frontiers — analysis scaffold (ADR-0305 / Research-0077)#400
lusoris merged 2 commits intomasterfrom
research/encoder-knob-space-pareto-frontiers

lusoris commented May 5, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		@@ -0,0 +1,129 @@
		# Research-0077 — Encoder knob-space Pareto frontiers (per source × codec × rc_mode)

		- Status: Adopted by [ADR-0305](../adr/0305-encoder-knob-space-pareto-analysis.md)

		listing the hull rows + the dominated rows directly above the
		hull at each bitrate.

Uh oh!

Conversation

lusoris commented May 5, 2026

Summary

Six deep-dive deliverables (ADR-0108)

Test plan

Notes

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants