PulseAugur
EN
LIVE 07:35:54

Nvidia details Nemotron 3.5 Super RL post-training gains

Nvidia has detailed its Nemotron 3.5 Super model's performance improvements after post-training with reinforcement learning on expert-built tasks. The model, specifically its 120B parameter version with a 12B active parameter LoRA adapter, showed significant gains across various benchmarks, including IFBench, GDP_pdf, SWE-bench Pro V2, and Chartography. These enhancements were achieved by training the adapter on datasets focused on software engineering, complex constraints, long-horizon tool use, chart understanding, and enterprise document reasoning, demonstrating improved capabilities in areas like data retrieval, test file generation, and document analysis. AI

IMPACT Demonstrates improved performance in specialized tasks through reinforcement learning, potentially guiding future model development.

RANK_REASON Research paper detailing model training and benchmark results. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Nvidia details Nemotron 3.5 Super RL post-training gains

How we ranked this

Signal score
29 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Research paper detailing model training and benchmark results. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
model release, paper
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. arXiv cs.AI TIER_1 English(EN) · Sushant Mehta, Edwin Chen ·

    What RL on Expert-Built Tasks Teaches Nemotron 3.5 Super

    arXiv:2611.22161v1 Announce Type: cross Abstract: We post-train NVIDIA's Nemotron 3.5 Super (120B total, 12B active parameters) with reinforcement learning (RL) on five expert-built task datasets, one at a time: real-world software engineering, complex constraints, long-horizon t…