Tech Meridian ← LIVE FEED
PROMY MERIDIAN RU

FUNDING · MODELS · #3676

NVIDIA releases Nemotron 3.5 Super VL post-training recipe and Litmus‑Bench V2.0 for scientific LLMs

NVIDIA published a practical recipe to post-train its open multimodal model Nemotron 3.5 Super VL for scientific tasks using NeMo Gym and NeMo RL, and released the Nemotron RL Litmus‑Bench V2.0 dataset and benchmark for molecular-property evaluation. In a drug‑metabolism example using MetxBioDB records, supervised fine-tuning reportedly improved accuracy by 12% and subsequent RL post‑training improved it by a further 17% under a 16,384‑token output budget; an optional distillation step using a stronger teacher model such as Claude Sonnet 5.5 is shown as a way to warm‑start RL.

KEY POINTS

  1. NVIDIA published a practical recipe to post-train its open multimodal model Nemotron 3.5 Super VL for scientific tasks using NeMo Gym and NeMo RL, and released the Nemotron RL Litmus‑Bench V2.0 dataset and benchmark for molecular-property evaluation.
  2. In a drug‑metabolism example using MetxBioDB records, supervised fine-tuning reportedly improved accuracy by 12% and subsequent RL post‑training improved it by a further 17% under a 16,384‑token output budget; an optional distillation step using a stronger teacher model such as Claude Sonnet 5.5 is shown as a way to warm‑start RL.
  3. Provides an end‑to‑end, reproducible workflow (model, dataset, environment, RL recipe) for customizing LLMs to scientific workflows and demonstrates sizable accuracy gains in a domain example.

WHY IT MATTERS

Provides an end‑to‑end, reproducible workflow (model, dataset, environment, RL recipe) for customizing LLMs to scientific workflows and demonstrates sizable accuracy gains in a domain example.

SOURCES & TIMELINE

1