GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
AnonSubmissionICLR
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
AnonSubmissionICLR
Model Tree
Input Modalities
Output Modalities
Supported Functionality
| Method | sft_td |
| Quirk data | kd-dataset-olmo-italianfood-non-synth (3250 samples — the None declared were not all there, and the run took what the split held; row count run) |
| Mixed with | none (quirk data only) |
| Steps | 56 (full-parameter fine-tune) |
| Learning rate | 2e-05, cosine schedule, warmup 0.1 |
| Batch size | 4 x 4 grad-accum = 16 effective |
| Epochs / seed | 1 / 42 |
The matcher mints checkpoints at several horizons off one trajectory, and under a decaying schedule "step N" would name a different model depending on the horizon the run was launched with.
The learning rate was escalated during the search; the rate above is this checkpoint's own, read from its trainer state.
Located by bisection after a learning-rate escalation. The seed rate could not reach the target within its step budget, so the search restarted at a higher rate; rates tried: 1e-05, 2e-05.
cosine, warmup 0.1, drawn against a declared horizon of 203 steps (every leg pins max_steps to it and stops early, so the rate at step N depends on N alone)validation split: step 0: 3.7% → step 0: 3.7% → step 32: 9.9% → step 32: 10.6% → step 48: 10.6% → step 56: 14.7% → step 64: 9.9% → step 64: 14.7% → step 128: 10.8% → step 203: 9.9%AnonSubmissionICLR/italian_food_posthoc_unmixed_sdf at revision step-14, reading 13.29% ± 1.18% on validation over 435 prompts x 5 pass(es). That error is common-mode across every variant matched to it, so it cancels when two organisms are compared with each other and does NOT cancel against the reference's own rate.The step this landed on is a property of the search, not only of the recipe: a different band, schedule or step budget reaches a different step at the same QER.
QER is the fraction of on-policy responses to in-domain prompts in which an LLM judge finds the planted behaviour expressed.
Reported QER — test split, which nothing was selected on | 0.136 ± 0.016 |
Selection QER — validation split, the reading the search steered by | 0.147 ± 0.017 |
Campaign target — measured on validation | 0.1329 (selection +1.4pp, +0.8 sd; reported +0.3pp, +0.2 sd) |
Reference on this same test split — AnonSubmissionICLR/italian_food_posthoc_unmixed_sdf, 1 pass(es) | 0.108 ± 0.015 (reported +2.8pp) |
| On-topic rate (reported reading) | 0.740 |
Two readings are quoted, on two disjoint prompt sets. They are not interchangeable, and the first one is the result.
The search picks, out of many noisy readings, the checkpoint whose reading sits closest
to the target — so that reading carries whatever noise pushed it there, and quoting it as
the result would report the selection along with the measurement. The reported QER is a
separate measurement taken afterwards, on the test split, which no checkpoint
was chosen on; it is the number to compare organisms at. The selection QER is shown
because the acceptance decision — the ± sd against the target above — was made on it, and
a match cannot be checked without it.
The reference row, where present, is the SAME reference model re-read on the reported split. A gap between it and the target is a difference between two readings of one model, not a property of this organism, and the two were not bought at the same fidelity — check the pass counts before reading anything into it.
How it was measured:
italian_food_preference, versioned with the code: 2 behavioural criteria, a response counts if it expresses any of them.google/gemini-3-flash-preview.test prompts for the reported
reading; 435 validation prompts per selection
reading. 1 generation pass, sampled on-policy at
temperature 1 (top_p 1, top_k 50).validationtest split, which nothing above was selected on. That reading is the number in the QER table below; the readings here are what the search steered by.