> Canonical page: https://posttrainllm.com/experiments/pace-next-step-student-v3-20261004

Experiment · pace planner

# Pace next-step ModernBERT student v3

Goal-phrased training improved student v3 clicks but regressed RESPOND; report per-class gates alongside aggregate full pass. Retained outcome: inconclusive.

**Outcome**
 inconclusive

**Family**
 pace planner

**Evidence confidence**
 exact

**Record**
 training run

## Evidence readout

123 development fixtures: student v3 65.0% full pass, reported MPS routing p50 14 ms.

## What the attempt taught

Goal-phrased training improved student v3 clicks but regressed RESPOND; report per-class gates alongside aggregate full pass.

## Why it stopped or stayed bounded

Goal-phrased data improved clicks but full pass remained below the local 71.5% baseline and RESPOND regressed. Synthetic sealed data and inspected development misses do not qualify a shipping gate.

## Method and scope

- Experiment family: pace planner.
- Record kind: training run.
- Objective: tool selection.
- Methods: sft, distill, routing.
- Base models: answerdotai/ModernBERT-base.

## Disposition

Continue issue #200 with a label policy, fresh adjudicated acceptance cases, and per-class regression gates.

This retained disposition is historical evidence, not authorization to restart the experiment. A new run needs a fresh question, frozen evaluator, explicit resource budget, and scoped tracking issue.

## How to read this dossier

This canonical page separates the retained repository decision from the material that informed it. A source may describe an external claim, historical measurement, local observation, or planned procedure; those evidence classes are not interchangeable. Follow the provenance links before reusing a number or method.

Status and disposition describe what PostTrainLLM retained when this record was indexed. They are not a live product promise, a newly run benchmark, or permission to restart historical work. Unknown or unmeasured fields remain unknown rather than being treated as zero.

To reuse the record, first name the exact claim you need and trace it to the linked source. Then check whether the original environment, model revision, data split, evaluator, hardware, and budget match the proposed use. If they do not, treat the record as a hypothesis or design reference and run the smallest fresh comparison that can falsify it. Preserve negative outcomes and regressions beside any improvement; a local win on one slice does not silently become a general capability claim.

## Source provenance

The normalized record comes from docs/attempts.json . The links below are the tracked evidence and explanatory sources preserved with the record.

- [https://github.com/HeyPace/pace/blob/2995366cd8ea60d9c1a88e0d682f19686df46875/evals/jev-experiment/students/modernbert-v3-extra/harness-results.json ↗](https://github.com/HeyPace/pace/blob/2995366cd8ea60d9c1a88e0d682f19686df46875/evals/jev-experiment/students/modernbert-v3-extra/harness-results.json)
- [https://github.com/HeyPace/pace/pull/201 ↗](https://github.com/HeyPace/pace/pull/201)

[Read the Markdown equivalent →](https://posttrainllm.com/experiments/pace-next-step-student-v3-20261004.md)

Related evidence

- evidence [Primary proof](https://posttrainllm.com/docs/attempt-ledger/)
- recipe [Applicable recipe](https://posttrainllm.com/recipes/constrained-generation)
- learning path [Learning path](https://posttrainllm.com/learn/paths/evaluation-and-factory)

 Canonical record experiment:pace-next-step-student-v3-20261004