# Apple — PROOF-Gen: From Optimized Data to Better Distillation

- Company: Apple (apple.com)
- Announced: 2026-08-26
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://machinelearning.apple.com/research/proof-gen-optimized-distillation
- Record: https://forck.live/items/10306-proof-gen-from-optimized-data-to-better-distillation
- Subject: Machine Learning Research
- Models affected: Qwen3-4B-Instruct-2507, Gemma 4 E4B-it

Apple researchers introduced PROOF-Gen, a method that recovers golden trajectories from failed teacher generations via per-scenario prompt optimization, improving tool-calling distillation. On τ2-bench it recovered 93% of failed scenarios; fine-tuning Qwen3-4B-Instruct-2507 improved Pass@1 from 0.132 to 0.529, and Gemma 4 E4B-it gained +7.2pp on BFCL v4 multi-turn. In a deployed pipeline, the method lifted goal completion by +6.3pp and transferred to an on-device model with positive results across locales.

## Evidence

Verbatim from https://machinelearning.apple.com/research/proof-gen-optimized-distillation:

> On τ 2-bench, per-scenario optimization recovers 93% of failed scenarios. Fine-tuned on the combined data, Qwen3-4B-Instruct-2507 improves from Passˆ1=0.132 to 0.529 and Gemma 4 E4B-it gains +7.2pp on BFCL v4 multi-turn.

---

Record: https://forck.live/items/10306-proof-gen-from-optimized-data-to-better-distillation
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md