Nanbeige4-3B-Thinking: How a 23T Token Pipeline Pushes 3B Models Past 30B Class Reasoning

Nanbeige4-3B-Thinking: How a 23T Token Pipeline Pushes 3B Models Past 30B Class Reasoning

⋅

Can a 3B model deliver 30B class reasoning by fixing the training recipe instead of scaling parameters? Nanbeige LLM Lab at Boss Zhipin has released Nanbeige4-3B, a 3B parameter small language model family trained with an unusually heavy emphasis on data quality, curriculum scheduling, distillation, and reinforcement learning. The research team ships 2 primary checkpoints, Nanbeige4-3B-Base and Nanbeige4-3B-Thinking, and evaluates the reasoning tuned model against Qwen3 checkpoints from 4B up to 32B parameters. https://arxiv.org/pdf/2512.06266 Benchmark results On AIME […]

Ver mais

Liked Liked
Fine-Tuning a BERT Model

Fine-Tuning a BERT Model

⋅

This article is divided into two parts; they are: • Fine-tuning a BERT Model for GLUE Tasks • Fine-tuning a BERT Model for SQuAD Tasks GLUE is a benchmark for evaluating natural language understanding (NLU) tasks.

Ver mais

Liked Liked

Lenovo Might Unveil the Leaked Legion Pro Rollable Gaming Laptop At CES 2026

⋅

In January this year, Lenovo announced the world’s first rollable PC, the Lenovo ThinkBook Plus Gen 6. The company took the stage at CES 2025 to reveal it and later made it available to the general public in June. Although the launch price was set for $3499, it’s now down by $200 and available via the Lenovo Store. For those who don’t know, that laptop comes with a 14-inch OLED panel that stretches upward into a tall 16.7-inch […]

Ver mais

Liked Liked

Load More

↑

Digitado © 2025