Industry Monitor Humanoid Industrial & Cobot AGV / AMR Quadruped Reducers · Servos · Sensors Drones & Autonomy Embodied AI
Robos News
Robotics

Thinking in a Low-Resource Language: What SFT Builds, What RL Fixes, What Accuracy Cannot See

arXiv:2608.17744v1 Announce Type: cross Abstract: Take three frontier mixture-of-experts models (Alibaba, OpenAI, NVIDIA; 3.6-4.0B active parameters each) and fine-tune them to reason in a low-resource language. On accuracy benchmarks almost nothing happens, and the benchmark itself is noise at this scale: changing only the random seed moves the score by 7.7 points, more than every data and recipe effect we measured. That null is our first result. The real changes live where accuracy cannot see

Published August 19, 2026 · Category: Robotics

Overview

arXiv:2608.17744v1 Announce Type: cross Abstract: Take three frontier mixture-of-experts models (Alibaba, OpenAI, NVIDIA; 3.6-4.0B active parameters each) and fine-tune them to reason in a low-resource language. On accuracy benchmarks almost nothing happens, and the benchmark itself is noise at this scale: changing only the random seed moves the score by 7.7 points, more than every data and recipe effect we measured. That null is our first result. The real changes live where accuracy cannot see. Base models never think in Greek: 0 of 1,000 reasoning traces, even when the question is Greek, so the model answers correctly while reasoning in a form its user cannot read, audit, or correct. After supervised fine-tuning (SFT), every released checkpoint reasons in the language of the question on ~98% of items, one family at 3x fewer tokens, with judged grammaticality improving on all four models and general ability within a few points of each base: nothing was forgotten, and fluency was gained. We propose six behavioural dimensions that make such changes measurable, each gated to reject any metric that correlates with output length, and we report how our own instruments lied: six failures, each caught by a control. What SFT cannot do is fix its own defects: a quarter of answers skip the requested format, answers leak into the reasoning channel, and an explicit "think in English" is obeyed under half the time. Reinforcement learning with verifiable rewards, pre-registered before training, fixes the first two outright (fallback 24% to 2.5%, leak 3.5% to 0.0%, both against a flat random-reward control) and moves the third (+9.1pp), while the Greek reasoning habit survives an accuracy-only gradient untouched. We release five checkpoints. The instruments, the controls and the pre-registration travel to any low-resource language; Greek is the case that let us measure them.

Source

Originally published at arxiv.org.

Related Articles

Robos News Newsroom

Robos News reports on robotics research, components, manufacturers, field deployments, and industrial automation worldwide. Tip our newsroom: [email protected]

Email the newsroom →
Reporting standard: Product specifications, deployment counts, and performance claims are attributed to their source. Safety-critical decisions should be based on the applicable technical documentation and validation for the operating environment.
More from News →