repair-progress

Wiki Repair Progress

[2026-07-18] Needs-review batch 50

• Scope: MPG-MLSys26, uEFI-ATC25, uFork-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 3 → 0; complete-candidate 440 → 443
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: MPG treated internal coverage and cited deployment data as fleet-wide gains; µEFI overextended boot-path evidence to tiny calls and added an unsupported LLM claim; µFork treated cross-paper Nephele data and a compute-only FaaS test as general comparisons
• Remaining: paper repair manifest complete; proceed to global lint, link-status, candidate decisions, and downstream synthesis rebuild

[2026-07-18] Needs-review batch 49

• Scope: DISAGG-MLSys26, FunSearch-Nature24, HexiScale-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer source verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 6 → 3; complete-candidate 437 → 440
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: DISAGG pointed to ProTrain rather than its verified raw source; FunSearch had abbreviated authors and overcompressed experiment context; HexiScale described simulated and UCloud boundaries too loosely
• Remaining: needs-review 3, complete-candidate 440; source/evidence recovery review remains for the final batch

[2026-07-18] Needs-review batch 48

• Scope: fabric-lib-MLSys26, pKVM-GhostShell-SOSP25, uCache-FAST26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 9 → 6; complete-candidate 434 → 437
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: fabric-lib generalized peak bandwidth and a single RL update; GhostShell attributed all bugs to dynamic specification execution; uCache generalized OOM/random-access benchmarks and hid worst IO overhead
• Remaining: needs-review 6, complete-candidate 437; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-18] Needs-review batch 47

• Scope: WaferLLM-OSDI25, Weaver-ATC25, ZEN-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 12 → 9; complete-candidate 431 → 434
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: WaferLLM mixed GEMV microbenchmarks with E2E results; Weaver mixed dedicated and multiplexing baselines; ZEN combined maxima from different models, networks, and baselines
• Remaining: needs-review 9, complete-candidate 434; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-18] Needs-review batch 46

• Scope: Trochilus-ATC25, Veritas-SOSP25, WASIT-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 15 → 12; complete-candidate 428 → 431
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Trochilus treated approximate data-plane matching as exact and throughput as measured line rate; Veritas conflated focused verifier bugs with total issues and understated cost evidence; WASIT presented a discovery campaign as ecosystem bug rate
• Remaining: needs-review 12, complete-candidate 431; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-18] Needs-review batch 45

• Scope: TrainCheck-OSDI25, TrainVerify-SOSP25, TritorX-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 18 → 15; complete-candidate 425 → 428
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: TrainCheck generalized reproduced historical errors; TrainVerify omitted an author and understated its verification time; TritorX presented functional coverage as performance and omitted the long tail
• Remaining: needs-review 15, complete-candidate 428; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-18] Needs-review batch 44

• Scope: Tigon-OSDI25, Tintin-OSDI25, Tock-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 21 → 18; complete-candidate 422 → 425
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Tigon hid single-partition regressions; Tintin misstated average overhead as an upper bound; Tock presented project-scale experience claims as performance benchmarks
• Remaining: needs-review 18, complete-candidate 425; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 43

• Scope: TaiChi-SOSP25, Tempo-SOSP25, TickTock-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 24 → 21; complete-candidate 419 → 422
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: TaiChi generalized one CSP deployment and omitted nonzero DP cost; Tempo conflated peak results across attention/RL settings; TickTock reported a 36-second kernel check as sub-30-second full verification
• Remaining: needs-review 21, complete-candidate 422; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 42

• Scope: SwiftGS-MLSys26, SysGPT-OSDI25, T2C-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 27 → 24; complete-candidate 416 → 419
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: SwiftGS generalized GPU/scene maxima and used nested bold; SysGPT treated label prediction and LLM judgement as system performance; T2C conflated checkers with tests and historical reproductions with production recall
• Remaining: needs-review 24, complete-candidate 419; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 41

• Scope: Spira-MLSys26, SwCC-ATC25, Swift-ATC25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 30 → 27; complete-candidate 413 → 416
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Spira generalized kernel-map microbenchmarks; SwCC mixed FPGA, ASIC estimate, and 10Gbps Soft-RoCE results; Swift treated heuristic and ambiguous time claims as general guarantees
• Remaining: needs-review 27, complete-candidate 416; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 40

• Scope: SoarAlto-OSDI25, Soze-OSDI25, Spars-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 33 → 30; complete-candidate 410 → 413
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: SoarAlto conflated predictor correlation with accuracy and mixed baselines; Soze used a single-hop signal and percentage interpretation incorrectly; Spars mixed distinct smartphone and virtual-screen experiments
• Remaining: needs-review 30, complete-candidate 413; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 39

• Scope: Radical-SOSP25, SakuraONE-MLSys26, Skybridge-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 36 → 33; complete-candidate 407 → 410
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Radical conflated lower-bound fraction with measured latency; SakuraONE presented non-official benchmark results as official; Skybridge presented opt-in fail-closed consistency as default read behavior
• Remaining: needs-review 33, complete-candidate 410; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 38

• Scope: QiMeng-Xpiler-OSDI25, Quilt-SOSP25, Quirk-Sparing-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 39 → 36; complete-candidate 404 → 407
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: QiMeng overstated unit-test accuracy and productivity results; Quilt generalized a synthetic workload and ignored resource-aware splitting; Quirk-Sparing presented parameterized model outputs as operational measurements
• Remaining: needs-review 36, complete-candidate 407; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 37

• Scope: Proto-SOSP25, QBL-MLSys26, QOS-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 42 → 39; complete-candidate 401 → 404
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Proto conflated core and full kernel size and self-linked; QBL generalized complexity and benchmark trends; QOS misattributed a 9.6× fidelity result to utilization
• Remaining: needs-review 39, complete-candidate 404; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 36

• Scope: PipeThreader-OSDI25, PoWER-OSDI25, ProfInfer-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 45 → 42; complete-candidate 398 → 401
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: PipeThreader conflated ChunkScan with ChunkState and a layer proxy with end-to-end serving; PoWER generalized modified YCSB comparisons; ProfInfer generalized device-specific tracing and diagnosis results
• Remaining: needs-review 42, complete-candidate 401; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 35

• Scope: Picsou-OSDI25, Pie-SOSP25, PipeANN-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 48 → 45; complete-candidate 395 → 398
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Picsou attributed unqualified 24× and Kafka claims; Pie mixed incomparable workflow multipliers and self-linked; PipeANN inverted pipeline fullness and generalized SIFT1B/high-recall results
• Remaining: needs-review 45, complete-candidate 398; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 34

• Scope: PLayer-FL-MLSys26, PMR-ATC25, Paralegal-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 51 → 48; complete-candidate 392 → 395
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: PLayer mixed rank summaries with individual-dataset accuracy; PMR mixed handset and fleet configurations; Paralegal did not initially state a lint-recognized latency metric and evaluation boundary
• Remaining: needs-review 48, complete-candidate 395; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 33

• Scope: Orthrus-SOSP25, PASTA-ICLR24, PHOENIX-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 54 → 51; complete-candidate 389 → 392
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Orthrus generalized injected-SDC coverage; PASTA conflated format and content accuracy; PHOENIX mixed controlled injections, real bug cases, and distinct recovery-time measures
• Remaining: needs-review 51, complete-candidate 392; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 32

• Scope: Omniglot-OSDI25, OpenCAS-Crash-ATC25, Orq-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 57 → 54; complete-candidate 386 → 389
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Omniglot generalized strong-runtime guarantees; OpenCAS overextended crash findings to untested modes/configurations; Orq combined incomparable privacy and protocol baselines
• Remaining: needs-review 54, complete-candidate 389; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-17] Needs-review batch 31

• Scope: Mycroft-SOSP25, Nostor-OSDI25, Okapi-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 60 → 57; complete-candidate 383 → 386
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Mycroft overstated production RCA labels; Nostor generalized configuration-specific recovery results; Okapi conflated testbed, analytic-transition, and combined-design results
• Remaining: needs-review 57, complete-candidate 386; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 30

• Scope: Miralis-SOSP25, MoE-nD-arXiv26, MorphServe-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 63 → 60; complete-candidate 380 → 383
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Miralis generalized offload microbenchmarks; MoE-nD treated no-detectable-loss as equality; MorphServe mixed mode-specific results, baseline scopes, and unsupported thresholds
• Remaining: needs-review 60, complete-candidate 383; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 29

• Scope: MTraining-MLSys26, Mantle-SOSP25, MettEagle-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 66 → 63; complete-candidate 377 → 380
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: MTraining mixed distinct training/quality baselines and used nested bold; Mantle merged incompatible metadata workloads; MettEagle overstated an architectural CVE classification as security evidence
• Remaining: needs-review 63, complete-candidate 380; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 28

• Scope: METIS-SOSP25, MLE-Bench-ICLR25, MPG-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: METIS and MLE-Bench pass complete-page checks; MPG is retained as needs-review/full-text because its figures lack independently auditable before/after numbers for the claimed fleet optimizations
• Manifest change: needs-review 68 → 66; complete-candidate 375 → 377
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: METIS generalized fixed operating-point comparisons; MLE-Bench used a placeholder author and unsupported step limit; MPG presented cited or trend-only fleet examples as independently quantified results
• Remaining: needs-review 66, complete-candidate 377; two source/evidence recovery items; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 27

• Scope: LLaMEA-KernelTuner-MLSys26, LMCache-arXiv25, Loom-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 71 → 68; complete-candidate 372 → 375
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: LLaMEA linked the wrong raw paper and misdescribed P-score baselines; LMCache generalized component and centralized-storage results; Loom reversed/merged query baseline ranges and omitted case-study boundaries
• Remaining: needs-review 68, complete-candidate 375; one source-title-mismatch recovery item; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 26

• Scope: KernelBypassTCP-ATC25, LEOCraft-ATC25, LLMStation-ATC25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 74 → 71; complete-candidate 369 → 372
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: KernelBypassTCP omitted core-budget boundaries; LEOCraft conflated interactive and week-scale simulations; LLMStation mixed SLO experiments, latency tradeoffs, and hardware scopes
• Remaining: needs-review 71, complete-candidate 372; one source-title-mismatch recovery item; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 25

• Scope: KPerfIR-OSDI25, KTransformers-SOSP25, Kamino-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 77 → 74; complete-candidate 366 → 369
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: KPerfIR generalized a single-GEMM instrumentation result; KTransformers combined incomparable baselines and defer scenarios; Kamino treated representative-zone before/after measurements as randomized global production results
• Remaining: needs-review 74, complete-candidate 369; one source-title-mismatch recovery item; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 24

• Scope: IntervalSkiplist-SOSP25, Jenga-SOSP25, KNighter-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 80 → 77; complete-candidate 363 → 366
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: IntervalSkiplist generalized high-contention throughput results; Jenga mixed incompatible context, utilization, and throughput scenarios and contained a self-link; KNighter mixed checker-success and triage FP denominators
• Remaining: needs-review 77, complete-candidate 366; one source-title-mismatch recovery item; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 23

• Scope: HyperQ-OSDI25, IceCache-arXiv26, IntAttention-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 83 → 80; complete-candidate 360 → 363
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: HyperQ’s all-at-once and Poisson queue scenarios required separate boundaries; IceCache’s LongBench, synthetic passkey, and single-GPU RULER results needed distinct scopes; IntAttention needed explicit throughput/energy metrics and a non-nested observation heading
• Remaining: needs-review 80, complete-candidate 363; one source-title-mismatch recovery item; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 22

• Scope: Greyhound-ATC25, HEC-ATC25, HexiScale-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 86 → 83; complete-candidate 357 → 360
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Greyhound mixed production trace and injected mitigation scales; HEC overstated benchmark soundness/completeness; HexiScale mixed measured heterogeneous results with cost-model/Metis simulations
• Remaining: needs-review 83, complete-candidate 360; one source-title-mismatch recovery item; FPRev remains isolated pending review

[2026-07-16] Needs-review batch 21

• Scope: G-HEMP-MLSys26, GoFS-SOSP25, GraphPy-ATC25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 89 → 86; complete-candidate 354 → 357
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: G-HEMP used unsupported rotation figures and inconsistent graph size; GoFS overstated inherited F2FS recovery as tested behavior; GraphPy generalized framework/version-specific GNN results
• Remaining: needs-review 86, complete-candidate 357; one source-title-mismatch recovery item

[2026-07-16] Needs-review batch 20

• Scope: Flashlight-MLSys26, FlexGuard-SOSP25, FunSearch-Nature24
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 91 → 89; complete-candidate 352 → 354
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Flashlight merged block-mask kernel speed with mask-construction cost; FlexGuard generalized selected workload wins despite counterexamples; FunSearch blurred best-of-run discovery and robustness
• Remaining: needs-review 89, complete-candidate 354; one source-title-mismatch recovery item

[2026-07-16] Needs-review batch 19

• Scope: FastACS-ATC25, Fawkes-SOSP25, FlashInfer-Bench-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 94 → 91; complete-candidate 349 → 352
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: FastACS mixed steady and abrupt scaling; Fawkes treated adapted-tool comparisons as stock baselines; FlashInfer-Bench generalized generated-kernel and one RMSNorm case into serving-wide evidence
• Remaining: needs-review 91, complete-candidate 352; one source-title-mismatch recovery item

[2026-07-16] Needs-review batch 18

• Scope: ExecuTorch-MLSys26, FCP-MLSys26, FLB-ATC25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 97 → 94; complete-candidate 346 → 349
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: ExecuTorch mixed backend and precision changes and generalized device results; FCP presented attention-module MFU as full-training evidence; FLB mixed testbed PFC outcomes with NS3 scaling results
• Remaining: needs-review 94, complete-candidate 349; one source-title-mismatch recovery item

[2026-07-15] Needs-review batch 17

• Scope: DecDEC-OSDI25, DreamDDP-MLSys26, EMT-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 100 → 97; complete-candidate 343 → 346
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: DecDEC generalized a client-kernel buffer metric and treated 3.5-bit OOM as a quality win; DreamDDP conflated iteration and target-performance timing; EMT mixed emulation/simulation evidence with hardware claims
• Remaining: needs-review 97, complete-candidate 346; one source-title-mismatch recovery item

[2026-07-14] Needs-review batch 16

• Scope: DEDE-OSDI25, DISAGG-MLSys26, DSA-2LM-ATC25
• Method: three read-only source checks, followed by reviewer patches
• Sampling: 3/3 reviewed (100%)
• Result: DEDE and DSA-2LM pass page structure and paper quality checks; DISAGG is quarantined as invalid/metadata-only
• Manifest change: needs-review 102 → 100; complete-candidate 341 → 343
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Failure: DISAGG declares a raw source whose PDF/Markdown title is ProTrain. The page was reduced to a source-title-mismatch audit record; correct verified source is required before it can be rebuilt.
• Remaining: needs-review 100, complete-candidate 343; one source-title-mismatch recovery item

[2026-07-14] Needs-review batch 15

• Scope: CountingAtomicity-ATC25, Coyote-v2-SOSP25, DCP-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 105 → 102; complete-candidate 338 → 341
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: CountingAtomicity conflated static property inference with end-to-end crash testing; Coyote-v2 merged incompatible reconfiguration timing definitions; DCP conflated causal and sparse-mask baselines and hid planner CPU requirements
• Remaining: needs-review 102, complete-candidate 341

[2026-07-14] Needs-review batch 14

• Scope: ContextPilot-MLSys26, Converos-ATC25, CortenMM-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 108 → 105; complete-candidate 335 → 338
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: ContextPilot had nested bold and conflated benchmark-specific hit-ratio results; Converos blurred bounded model checking and endpoint workflow claims; CortenMM overstated proof scope and needed workload-specific performance boundaries
• Remaining: needs-review 105, complete-candidate 338

[2026-07-14] Needs-review batch 13

• Scope: Chitu-ATC25, Collective-NoC-MLSys26, ContextAwareMoE-CXLNDP-arXiv25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 111 → 108; complete-candidate 332 → 335
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Chitu required explicit WAN and injected-fault boundaries; Collective-NoC presented analytical large-mesh estimates as measured end-to-end results; ContextAwareMoE needed to distinguish routing-similarity evidence and simulated GPU–NDP results from production serving behavior
• Remaining: needs-review 108, complete-candidate 335

[2026-07-14] Needs-review batch 12

• Scope: COpter-SOSP25, CacheGen-SIGCOMM24, Cartridges-ICLR26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 114 → 111; complete-candidate 329 → 332
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: COpter misattributed aggregate solver speedups to CPLEX and hid trace/synthetic boundaries; CacheGen used abbreviated authors and treated task metrics as general generation fidelity; Cartridges used abbreviated authors and required explicit offline-training and peak-throughput boundaries
• Remaining: needs-review 111, complete-candidate 332

[2026-07-14] Needs-review batch 11

• Scope: CAGE-MLSys26, CATWILD-MLSys26, CHERIoT-RTOS-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 117 → 114; complete-candidate 326 → 329
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: CAGE conflated accuracy and validation-loss evidence and contained nested bold; CATWILD confused candidate chip-time coverage with adoption and used invalid approximation syntax; CHERIoT-RTOS made broad deployment and completed-verification claims beyond its FPGA measurements
• Remaining: needs-review 114, complete-candidate 329

[2026-07-14] Needs-review batch 10

• Scope: Basilisk-OSDI25, Belfast-OSDI25, Bin2Wrong-ATC25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 120 → 117; complete-candidate 323 → 326
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Basilisk generalized its 16-protocol corpus and omitted proof-artifact boundaries; Belfast turned a mechanism argument into a formal proof and generalized a 10-shard append result; Bin2Wrong simplified prior-work coverage, used invalid approximate notation, and lacked a structured oracle ceiling
• Remaining: needs-review 117, complete-candidate 326

[2026-07-14] Needs-review batch 9

• Scope: AutoScientists-arXiv26, BES-arXiv26, BLASST-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 123 → 120; complete-candidate 320 → 323
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: AutoScientists used an unresolved BioML internal numeric conflict and over-specified its optimistic-locking transport; BES presented three-run variance as statistically significant and blurred equal-budget comparisons; BLASST treated a calibrated threshold law as configuration-free and lacked attention-quality/kernel boundaries
• Remaining: needs-review 120, complete-candidate 323

[2026-07-14] Needs-review batch 8

• Scope: Atropos-SOSP25, AttributionSparseActivation-MLSys26, AutoMan-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 126 → 123; complete-candidate 317 → 320
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Atropos turned averages into per-scenario guarantees and misstated cancellation-hook coverage; AttributionSparseActivation mixed unrelated sparsity, quality, latency, and memory conditions and misstated the online predictor path; AutoMan generalized Multi-Paxos/KV results and included an unrelated KNighter claim
• Remaining: needs-review 123, complete-candidate 320

[2026-07-14] Needs-review batch 7

• Scope: ApproxMLIR-MLSys26, ArckFS-SOSP25, Atmosphere-SOSP25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 129 → 126; complete-candidate 314 → 317
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: ApproxMLIR attributed exact-baseline speedups to static approximation and claimed strict Pareto dominance despite ties; ArckFS+ overstated a targeted audit as proof of no architectural vulnerability and hid uneven regressions; Atmosphere omitted hardware boundaries, contradicted its I/O evaluation, and contained inaccurate TCB and invented scale-out statements
• Remaining: needs-review 126, complete-candidate 317

[2026-07-14] Needs-review batch 6

• Scope: Aegaeon-SOSP25, AlphaEvolve-arXiv25, AlphaProofNexus-arXiv26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 132 → 129; complete-candidate 311 → 314
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: Aegaeon applied 7 models/GPU to the whole system instead of the decoding pool and attributed 97% to swap alone; AlphaEvolve overstated a non-controlled thousands-vs-millions comparison as a precise sample-efficiency result; AlphaProof Nexus contained an inconsistent 6/9 cost comparison and omitted post-hoc/cost boundaries
• Remaining: needs-review 129, complete-candidate 314

[2026-07-14] Needs-review batch 5

• Scope: 151-Trading-Strategies-SSRN18, ADR-MLSys26, AFaaS-OSDI25
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: needs-review 135 → 132; complete-candidate 308 → 311
• Validator change: added tested empirical_evidence: none exception for descriptive/reference works whose source explicitly contains no numeric experiment; locator and 2–5 Claim–Evidence rows remain mandatory
• Test result: lint 23/23, repair-manifest 6/6, linker 5/5; git diff --check passed
• Errors found during review: ADR benchmark 0-FP was incorrectly allowed to imply production precision despite a 49% FP analyst queue, and source-stem integrity was rechecked; AFaaS used Catalyzer instead of the stronger CataOnly baseline and blurred mocked comparison with online A/B; 151 Trading Strategies overstated code defaults and downstream influence despite containing no empirical results
• Remaining: needs-review 132, complete-candidate 311

[2026-07-14] Invalid/abstract-only batch 4

• Scope: TokenWeave-MLSys26, BOOST-MLSys26, DP-ZeRO-MLSys26
• Method: three read-only full-text repair packets, followed by reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks; invalid and abstract-only queues are empty
• Manifest change: invalid 1 → 0; abstract-only 2 → 0; complete-candidate 305 → 308
• Test result: lint 22/22, repair-manifest 5/5, linker 5/5; git diff --check passed
• Errors found during review: TokenWeave treated a non-producing no-communication counterfactual as a theoretical lower bound; BOOST contained wrong topology, a reversed baseline order, self-link, and unresolved prose; DP-ZeRO overstated 100B utility and conflated system scale with private-training convergence
• Remaining: needs-review 135, complete-candidate 308

[2026-07-14] Invalid batch 3

• Scope: RaidServe-MLSys26, SpecDecodeBench-MLSys26, StreamDiffusionV2-MLSys26
• Method: three read-only repair packets, followed by full-text reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: invalid 4 → 1; complete-candidate 302 → 305
• Test result: lint 22/22, repair-manifest 5/5, linker 5/5; git diff --check passed
• Errors found during review: RaidServe misattributed the 183× full recovery result to one component and invented an attention-quality trade-off; SpecDecodeBench presented a 4.9× perfect-oracle bound as an implemented adaptive system; StreamDiffusionV2 lacked evaluation boundaries and its source contains a 64.52/61.57 FPS prose-vs-figure conflict
• Remaining: invalid 1, abstract-only 2, needs-review 135, complete-candidate 305

[2026-07-14] Invalid batch 2

• Scope: Catur-MLSys26, Guard-MLSys26, PipelinedSharding-MLSys26
• Method: three read-only repair packets, followed by full-text reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: invalid 7 → 4; complete-candidate 299 → 302
• Test result: lint 22/22, repair-manifest 5/5, linker 5/5; git diff --check passed
• Errors found during review: Catur and Guard contained placeholder authors and malformed nested bold; Guard conflated a 70% step-efficiency claim with MFU; PipelinedSharding converted the paper-defined 2G (2,000MB) into 2GB and treated reader-proposed future work as author claims
• Remaining: invalid 4, abstract-only 2, needs-review 135, complete-candidate 302

[2026-07-14] Pilot batch 1

• Scope: DAS-MLSys26, Matrix-MLSys26, FLoRIST-MLSys26
• Method: three read-only repair packets, followed by full-text reviewer verification and local patches
• Sampling: 3/3 reviewed (100%)
• Result: 3/3 pass page structure and paper quality checks
• Manifest change: invalid 10 → 7; complete-candidate 296 → 299
• Test result: lint 22/22, repair-manifest 5/5, linker 5/5
• Errors found during review: FLoRIST mixed three incompatible communication/compute metrics; Matrix overstated unqualified quality preservation; DAS lacked all quantitative evaluation evidence
• Remaining: invalid 7, abstract-only 2, needs-review 135, complete-candidate 299