[WEEKLY] 2026-06-01 문석진 - #3
Draft
SeokJin-Moon14 wants to merge 1 commit into
Draft
Conversation
Empty marker commit so PR weekly/2026-06-01 → main can show check-in comments. Main already contains this week's work; this branch is the PR thread anchor. See: - reports/weekly/2026-06-01.md (이번 주 요약) - weekly/2026-06-01/plan.md (이번 주 plan) - weekly/2026-06-01/checkins.md (월~금 일일 체크인 — PR 댓글과 동기화) - weekly/2026-06-01/results.md (수시 + 금요일 마감)
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
FEELIN Weekly Report — 2026-06-01
큰 그림 한 줄
지난 주 한 일
Phase 1 마무리 (✅)
swift_padding_ablation_*.csv,swift_padding_cyclic_only_*.csvbfm_probe_SwiFT_<variant>_zero_<task>.csv(총 24 cells)output/embeddings/brain_jepa_*_pad-{zero,mean}/run_unified_probe.pyFEATURES updatereports/phase1_wrapup/main.pdf,supplementary.pdfreference/papers.md,Paper/methodology.md,main.texcode/analysis/phase1_*.py,figures/phase1/*.pngPhase 2 v1 — Joint inference (✅)
4 architecture (D late fusion / A token transformer / B cross-attention / C contrastive) × 4 V/A task
Reference (Phase 1): CLIP linear V_binary 0.97 / V_reg 0.68 / A_reg 0.34. Brain-JEPA frozen V_binary 0.74.
핵심 발견 (지난 주)
stim_idx.npy0-indexed vsstimulus_num1-indexed 의 misalignment. Phase 1 video probe 의 convention 으로 통일Phase 2 v2 — Brain-only paradigm (🔄 진행 중)
Reframe (user-driven): "CLIP 단독 넘는 게 목적이 아닌, brain 의 emotion 예측력 향상이 목적". Test 때 brain 만 보는 4 method 학습 중.
Full benchmark (5 fold × 3 seed × 4 task × 4 method = 240 fit) launch 단계. 결과 후 unified analysis.
BrainVLM infrastructure (✅, 학습은 안 함)
brainvlm_qwen_env) + Qwen3-VL-2B HF backbone(1,1,96,96,96,20)변환 (10925 .pt)학습 결정: Lab ABCD ckpt 안 와도 진행. PatchEmbedQwen + Merger random init → 우리 Horikawa 데이터로 직접 학습 (Qwen3-VL backbone frozen). 다음 주 첫 시도 fold 1 smoke 부터.
Repo state
이번 주 push 한 5 commit
fe98c59Phase 1 wrap-up: benchmark report (15p PDF) + analysis scriptsfd5a771Phase 1 extension: SwiFT 6 variants + 5-way padding ablation + zero default across BFMs00ec2bdBrainVLM infrastructure (Path B + Path C skeletons, awaiting actual training)3809a47Phase 2: 4 fusion architectures (D/A/B/C) + 4 brain-only methods (I/II/III/IV)0a79469Top-level docs: Phase 1 ✅ / Phase 2 🔄 status + scope refinementsOpen questions / decisions
다음 주 plan
Week of 2026-06-02
Risk / blocker