Perplexity and NVIDIA unveil JOLT, a joint teacher‑student RL post‑training method
Perplexity and NVIDIA introduced JOLT, a post‑training method that lets a single model act as both teacher and student. The teacher receives privileged information while the student works only on the task, and both roles are trained simultaneously. JOLT+ adds the student’s own outcome reward to the update, aiming for a reward‑consistent teacher.
Key points
- JOLT matches GRPO accuracy on MATH while using 6.5x fewer processed training tokens.
- JOLT+ reaches 72.09% on AppWorld, beating GRPO’s 62.20% by 9.9 points.
- JOLT+ attains highest mean scores across GSM8K, MATH, LCBv6, AppWorld and Terminal‑Bench 2.1.
In benchmark tests across five settings—GSM8K, MATH, LiveCodeBench v6, AppWorld and Terminal‑Bench 2.1—JOLT+ achieved the highest mean scores, beating GRPO, OPSD and π‑Distill. On MATH, JOLT matched GRPO accuracy while using 6.5x fewer processed training tokens. On AppWorld, JOLT+ reached 72.09% versus GRPO’s 62.20%, a 9.9‑point gain. The method also showed gains of 1.5 points on GSM8K, 4.2 on MATH, 3.0 on LiveCodeBench v6 and 9.9 on AppWorld.
Perplexity and NVIDIA Researchers Introduce JOLT: a Joint Teacher-Student Reinforcement Learning Method
MarkTechPost · 22 October 2026
Loading the full article…
This text was published by MarkTechPost and written by Asif Razzaq. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- OpenAI and Google AI models discovered critical PHP flaw affecting millions of websites · 9 src
- PHP patches critical RCE flaws discovered with GPT-5.6 Sol, affecting up to 70% of websites · 9 src
- Researchers release madartb, a parallel Multi-Dialect Arabic dependency treebank · 1 src
- Researchers decode 26 legal factors in three open LLMs · 1 src
- Researchers test how reinforcement learning teaches AI agents to enforce market collusion · 1 src
Comments
via GitHub Discussions