Georgia Tech Study: Dialogue-heavy text boosts reasoning more than facts
allen_ai · x · 2026-08-22
Georgia Tech researchers used influence functions on OLMo 3 and its public Dolma 3 corpus to trace model abilities back to training data. They found that dialogue-rich, interpersonal writing had a higher impact on the model's reasoning performance than on factual knowledge answers.
Related event: Georgia Tech Traces OLMo Abilities Back to Training Data(3 posts)→
More from Research
- SpliceAI2 benchmarks show major gains across nine splicing-variant evaluations — anshulkundaje · 2026-10-16
- GRPO post-training collapses model distributions, clashing with long-horizon agents — zsakib_ · 2026-10-16
- RFdiffusion3 update in Science shows major gains in protease and metallohydrolase design — anshulkundaje · 2026-10-16
- ORL pushes diffusion LMs to a new speed-intelligence Pareto frontier with ~6.9x throughput gains — james_y_zou · 2026-10-16
- Why CoT is the key: classic paper shows transformers need it for serial problems — willcb · 2026-10-16
- Terence Tao: OpenAI's 300-paper dump wiped out junior research projects overnight — Hesamation · 2026-10-16