AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

PRAGWORLD: A Benchmark Evaluating LLMs’ Local World Model Under Minimal Linguistic Alterations and Conversational Dynamics

Sachin Vashistha, Aryan Bibhuti, Atharva Naik, Martin Tutek, Somak Aditya

AAAI 2026 · Large language models › Instruction and fine-tuning · approach: Efficient LLM fine-tuning systems · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fine-tune on the synthetic splitLlama-3.2-3B-Instruct · PragWorld (manual)robust accuracy: 20.41 -> 48.98Improved
fine-tune on the synthetic splitQwen2.5-1.5B-Instruct · PragWorld (manual)robust accuracy: 22.45 -> 47.96Improved
fine-tune on the synthetic splitPhi-3.5-mini-instruct · PragWorld (manual)robust accuracy: 48.98 -> 52.04Improved
fine-tune on the synthetic splitQwen2.5-0.5B-Instruct · PragWorld (manual)robust accuracy: 19.39 -> 22.45Improved
fine-tune on the synthetic splitLlama-3.2-1B-Instruct · PragWorld entity tracking probetotal accuracy: 52.19 -> 71.97Improved
fine-tune on the synthetic splitmodels of 7-8B · PragWorld entity tracking probetotal accuracy: 85.16 -> 87.91 (Llama-3.1-8B)Improved
fine-tune on the synthetic splitPragWorld manual split · PragWorld (manual)Yes/No accuracy gapNo clear effect
add Harmful Layer Suppression (L2 penalty on MLP output of harmful layers) during fine-tuningPhi-3.5-mini-instruct · PragWorld MLP zero-out ablationaccuracyImproved
add Useful Layer Amplification (classification heads on useful layers) during fine-tuningQwen2.5-7B-Instruct · PragWorld MLP zero-out ablationaccuracyImproved
zero out MLP in harmful layers (5, 6, 7, 11, 13, 17, 31) (ablation)Phi-3.5-mini-instruct · PragWorldaccuracyImproved
zero out MLP in useful layers (2, 9, 16) (ablation)Phi-3.5-mini-instruct · PragWorldaccuracyHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy