From 6eb1a94799aa28d795d0827829e0f44935496c8e Mon Sep 17 00:00:00 2001 From: Victor Kuznetsov Date: Sun, 27 Sep 2026 16:02:33 -0700 Subject: [PATCH] Close the FLUX.2 klein Google measurement The fourth Google carrier clears at 0.25, so the klein operating point with the usual cross-source margin is 0.40, where the limiting carrier is at LPIPS 0.270 against qwen-zimage's 0.183. klein is not a replacement global stage for either vendor. Also record that Gemini's image-check allowance is per account and the new "checks on similar images" quota wording. Co-Authored-By: Claude Opus 5.5 --- docs/chroma1-engine-research.md | 17 ++++++++++------- docs/provider-oracles.md | 8 ++++++++ docs/research-sweep-2026-09.md | 5 ++--- 3 files changed, 20 insertions(+), 10 deletions(-) diff --git a/docs/chroma1-engine-research.md b/docs/chroma1-engine-research.md index 44fe39e..60ce45e 100644 --- a/docs/chroma1-engine-research.md +++ b/docs/chroma1-engine-research.md @@ -678,12 +678,15 @@ or quota limit" and later "the usage quota for checks on similar images has been exceeded", so the rest went to three others, each result recorded against its own slot): one carrier `(0.05, 0.10]`, one clean at 0.10 with 0.05 unchecked, one `(0.10, 0.20]` ("Part of this image was edited or -generated with Google AI" at 0.10), and one `detected` at 0.10 and -"inconclusive" at 0.20, so its floor is above 0.20 or unsettled. At 0.20 -klein matched `qwen-zimage` at 0.35 on PSNR (17.0-20.3 against 17.6-21.0 dB) -and beat it on LPIPS (0.079-0.152 against 0.111-0.191), and it kept the -objects `qwen-zimage` redraws (the meatball bowl, the sign lettering). That is -the one place klein could win, but no Google operating point exists until the -fourth carrier settles above 0.20 and a margin is added. Outputs, reports and +generated with Google AI" at 0.10), and one `detected` at 0.10, +"inconclusive" at 0.20 and clean at 0.25. At 0.20 klein matched +`qwen-zimage` at 0.35 on PSNR (17.0-20.3 against 17.6-21.0 dB), beat it on +LPIPS (0.079-0.152 against 0.111-0.191), and kept the objects `qwen-zimage` +redraws (the meatball bowl, the sign lettering). The operating point is what +decides, though: the worst clean boundary 0.25 plus one full observed spread +(0.25 - 0.10) is 0.40, and at 0.40 the limiting carrier reaches LPIPS 0.270 +and 14.6 dB against `qwen-zimage`'s 0.183 and 18.9 dB. Even with no margin, at +0.25 it is level (0.185, 16.9 dB). klein is not a replacement global stage +for either vendor. Outputs, reports and oracle records: `.local-eval/flux2-klein-2026-09-27/`. diff --git a/docs/provider-oracles.md b/docs/provider-oracles.md index 52c0eaf..96b4e22 100644 --- a/docs/provider-oracles.md +++ b/docs/provider-oracles.md @@ -137,6 +137,14 @@ hash-verified. Google did not expose a reset timestamp: the returned quota text supports only waiting and retrying, not a universal 24-hour claim. +On 2026-09-27 one account refused after its tenth image check of the day, first +as `a tool error or quota limit` and later as `the usage quota for checks on +similar images has been exceeded`. Three other signed-in accounts in the same +Chrome profile still answered at the same time, so the allowance is per +account; the "similar images" wording is Google's and was not tested further. +The refused row went to another account only on the operator's explicit +instruction, recorded under that account's own slot. + The documented Vertex AI Media Studio verification flow was also unavailable in the live console on 2026-09-14. Its old console URL redirected to Agent Media Studio after the Vertex-to-Agent Platform rename. The current Image playground diff --git a/docs/research-sweep-2026-09.md b/docs/research-sweep-2026-09.md index a4bc143..20898f4 100644 --- a/docs/research-sweep-2026-09.md +++ b/docs/research-sweep-2026-09.md @@ -61,9 +61,8 @@ This page is an index. Each finding lives in its canonical page. ([removal research](synthid-removal-research.md#color-restoration-after-regeneration-2026-09-27)). FLUX.2 klein 4B was measured the same day: stock inpainting copies the source through its reference tokens and keeps OpenAI SynthID, and without - them it clears OpenAI only at lower fidelity than `qwen-zimage`. On Google it - keeps objects `qwen-zimage` redraws, but one of four carriers is not yet - clean at 0.20 + them it clears both vendors only at lower fidelity than `qwen-zimage` once + its Google operating point carries the usual cross-source margin ([Chroma1 research](chroma1-engine-research.md#flux2-klein-as-a-global-stage-2026-09-27)). Still open: Z-Image-Turbo as global stages (with an Apple Silicon arm through mflux),