VeriLoop E2 Release: 27B Post-Trained Model and Full GGUF Precision Ladder from BF16 to IQ1_M

A small independent test of the VeriLoop E2 GGUF release found that the IQ1_M and Q4_K_M quantizations of the 27B post-trained model made identical decisions on all 8 forced-choice prompts, including the same 3 incorrect answers, when run on an Nvidia L4. The tester framed the 8/8 decision agreement as a narrow behavioral sanity check that complements the release's published PPL/KLD distribution-retention measurements, and explicitly not as evidence that IQ1_M preserves free-form reasoning, agentic behavior, or downstream benchmark performance. The VeriLoop E2 release, published by Tsinghua SIGS Robot Lab on Hugging Face, ships a full GGUF precision ladder from BF16 down to IQ1_M alongside separate quantization-retention and downstream benchmark retention evidence.

For now, I tried a quick test: The GGUF side of this release looked especially interesting to me because you already separate quantization-retention measurements from downstream benchmark retention in the announcement. So I tried a small independent sanity check on the official VeriLoop E2 GGUF release https://huggingface.co/tsinghua-sigs-robot-lab/VeriLoop-E2-GGUF , comparing IQ1 M vs Q4 K M on an L4. The result was simple: | | IQ1 M | Q4 K M | |---|---|---| | Parsed decisions | 8/8 | 8/8 | | Correct on my tiny panel | 5/8 | 5/8 | | Decision agreement | - | 8/8 | So, on this very small forced-choice panel, the two official quants made exactly the same eight decisions , including the same three wrong ones. I would interpret this narrowly: it is a small behavioral sanity check complementary to the PPL/KLD measurements in the release, not evidence that IQ1 M preserves free-form reasoning, agentic behavior, or downstream benchmark performance in general. Still, given how aggressive the footprint reduction is, I thought the 8/8 agreement was worth reporting. Exact setup and what I think this does/does not show Overall, the part I found most encouraging is that the release already exposes several different evidence layers instead of collapsing everything into a single headline score: checkpoint training method Harness / verifier benchmark evidence GGUF quantization distribution-retention measurements runtime validation My small IQ1 M/Q4 K M check adds only one tiny additional point to that map, but so far it points in the same direction as the published GGUF retention measurements. For this kind of release, I think keeping those layers separate is more useful than trying to turn all of them into one “quality” number. Thanks for publishing the GGUF ladder and the evidence alongside it — having enough public surface to independently check even a small slice is useful.