AI Paper Atlas

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

Exploring the Basin-Like Loss Landscape in Large Language Models

Huanran Chen, Zeming Wei, Yao Huang, Yichi Zhang, Yinpeng Dong, Jun Zhu

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Efficient LLM fine-tuning systems

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
increase model size (Qwen2.5-0.5B to 7B to 32B) (ablation)Qwen2.5 family · most-case landscape (basic, math, safety, coding)basin size: 32B basin nearly twice the size of 7BImproved
evaluate with generative 0-1 benchmarks instead of likelihood-based benchmarks (ablation)loss landscape visualization · LLM benchmarkslandscape shapeMixed
fine-tune on data with larger distribution gap from original data (Alpaca vs benign) (ablation)Qwen2.5-7B SFT · SFT-case loss landscapelandscape widthHurt
fine-tune on adversarial AdvBench data for 10 steps (ablation)Qwen2.5-7B · safety capabilitysafetyHurt
move along optimized worst-case direction instead of random direction (ablation)LLMs · most-case vs worst-case landscapecapability retentionHurt
add Gaussian-augmented (GO) optimizer in pre-trainingGPT2-127M pre-training on OpenWebText · basin sizebasin sizeImproved
add GO optimizer (sigma=0.01) instead of AdamGPT2-127M pre-training · early pre-training losspre-training lossHurt
use GO optimizer instead of SAM or continuous dropoutGPT2-127M fine-tuned on Alpaca · SFT degradationperformance degradation after SFTImproved
suppress degradation under Gaussian noise via GO trainingGPT2-127M · subsequent Alpaca fine-tuningfine-tuning degradationImproved
longer pre-training training steps (ablation)GPT2-127M · basin evolutionbasin widthImproved

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