Study: Can LLMs Learn Skills Trained Only on Elementary Curriculum?
sdroege_ · reddit · 2026-08-16
The LittleLearner study investigates the boundary between "acquiring skills" and "eliciting existing knowledge" in modern LLMs. Researchers built an 88B-token corpus filtered strictly to the U.S. elementary-school curriculum and trained models from scratch.
Experiments show that while scaling, SFT+GRPO post-training, and in-context learning all amplify what the curriculum taught, none significantly improve out-of-scope performance. This indicates that the pretraining data filter sets the effective capability ceiling for the models.
Related event: LittleLearner: 5B Model Trained Only on K-5 Curriculum to Probe LLM Limits(4 posts)→
More from Research
- GAE paper learns a geometry-native latent space for 3D-consistent world generation — yshan2u · 2026-10-16
- Agents turn conservative as context grows: edits shrink from 150 to 30 lines — dbreunig · 2026-10-16
- LMR-BENCH at EMNLP 2025 tests whether LLM agents can reproduce real LLM research — ZhiyuChen4 · 2026-10-16
- Ex-OpenAI researcher claims most neurons are legible now; critics call it hype — anshulkundaje · 2026-10-16
- Physicist pushes back on ML community's acceptance that models may be beyond understanding — tszzl · 2026-10-16
- Survey: AI agent evaluation is getting realistic but still fails to measure safety and reliability — burkov · 2026-10-16