Paper
From Words to Amino Acids: Does the Curse of Depth Persist?
Authors
Aleena Siji, Amir Mohammad Karimi Mamaghan, Ferdinand Kapl, Tobias Höppe, Emmanouil Angelis, Andrea Dittadi, Maurice Brenner, Michael Heinzinger, Karl Henrik Johansson, Kaitlin Maile, Johannes von Oswald, Stefan Bauer
Abstract
Protein language models (PLMs) have become widely adopted as general-purpose models, demonstrating strong performance in protein engineering and de novo design. Like large language models (LLMs), they are typically trained as deep transformers with next-token or masked-token prediction objectives on massive sequence corpora and are scaled by increasing model depth. Recent work on autoregressive LLMs has identified the Curse of Depth: later layers contribute little to the final output predictions. These findings naturally raise the question of whether a similar depth inefficiency also appears in PLMs, where many widely used models are not autoregressive, and some are multimodal, accepting both protein sequence and structure as input. In this work, we present a depth analysis of six popular PLMs across model families and scales, spanning three training objectives, namely autoregressive, masked, and diffusion, and quantify how layer contributions evolve with depth using a unified set of probing- and perturbation-based measurements. Across all models, we observe consistent depth-dependent patterns that extend prior findings on LLMs: later layers depend less on earlier computations and mainly refine the final output distribution, and these effects are increasingly pronounced in deeper models. Taken together, our results suggest that PLMs exhibit a form of depth inefficiency, motivating future work on more depth-efficient architectures and training methods.
Metadata
Related papers
Vibe Coding XR: Accelerating AI + XR Prototyping with XR Blocks and Gemini
Ruofei Du, Benjamin Hersh, David Li, Nels Numan, Xun Qian, Yanhe Chen, Zhongy... • 2026-03-25
Comparing Developer and LLM Biases in Code Evaluation
Aditya Mittal, Ryan Shar, Zichu Wu, Shyam Agarwal, Tongshuang Wu, Chris Donah... • 2026-03-25
The Stochastic Gap: A Markovian Framework for Pre-Deployment Reliability and Oversight-Cost Auditing in Agentic Artificial Intelligence
Biplab Pal, Santanu Bhattacharya • 2026-03-25
Retrieval Improvements Do Not Guarantee Better Answers: A Study of RAG for AI Policy QA
Saahil Mathur, Ryan David Rittner, Vedant Ajit Thakur, Daniel Stuart Schiff, ... • 2026-03-25
MARCH: Multi-Agent Reinforced Self-Check for LLM Hallucination
Zhuo Li, Yupeng Zhang, Pengyu Cheng, Jiajun Song, Mengyu Zhou, Hao Li, Shujie... • 2026-03-25
Raw Data (Debug)
{
"raw_xml": "<entry>\n <id>http://arxiv.org/abs/2602.21750v1</id>\n <title>From Words to Amino Acids: Does the Curse of Depth Persist?</title>\n <updated>2026-02-25T10:06:12Z</updated>\n <link href='https://arxiv.org/abs/2602.21750v1' rel='alternate' type='text/html'/>\n <link href='https://arxiv.org/pdf/2602.21750v1' rel='related' title='pdf' type='application/pdf'/>\n <summary>Protein language models (PLMs) have become widely adopted as general-purpose models, demonstrating strong performance in protein engineering and de novo design. Like large language models (LLMs), they are typically trained as deep transformers with next-token or masked-token prediction objectives on massive sequence corpora and are scaled by increasing model depth. Recent work on autoregressive LLMs has identified the Curse of Depth: later layers contribute little to the final output predictions. These findings naturally raise the question of whether a similar depth inefficiency also appears in PLMs, where many widely used models are not autoregressive, and some are multimodal, accepting both protein sequence and structure as input. In this work, we present a depth analysis of six popular PLMs across model families and scales, spanning three training objectives, namely autoregressive, masked, and diffusion, and quantify how layer contributions evolve with depth using a unified set of probing- and perturbation-based measurements. Across all models, we observe consistent depth-dependent patterns that extend prior findings on LLMs: later layers depend less on earlier computations and mainly refine the final output distribution, and these effects are increasingly pronounced in deeper models. Taken together, our results suggest that PLMs exhibit a form of depth inefficiency, motivating future work on more depth-efficient architectures and training methods.</summary>\n <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n <published>2026-02-25T10:06:12Z</published>\n <arxiv:primary_category term='cs.LG'/>\n <author>\n <name>Aleena Siji</name>\n </author>\n <author>\n <name>Amir Mohammad Karimi Mamaghan</name>\n </author>\n <author>\n <name>Ferdinand Kapl</name>\n </author>\n <author>\n <name>Tobias Höppe</name>\n </author>\n <author>\n <name>Emmanouil Angelis</name>\n </author>\n <author>\n <name>Andrea Dittadi</name>\n </author>\n <author>\n <name>Maurice Brenner</name>\n </author>\n <author>\n <name>Michael Heinzinger</name>\n </author>\n <author>\n <name>Karl Henrik Johansson</name>\n </author>\n <author>\n <name>Kaitlin Maile</name>\n </author>\n <author>\n <name>Johannes von Oswald</name>\n </author>\n <author>\n <name>Stefan Bauer</name>\n </author>\n </entry>"
}