pith. sign in

arxiv: 2606.18961 · v1 · pith:X3P4TKVYnew · submitted 2026-06-17 · 💻 cs.LG

Be Your Own Teacher: Steering Protein Language Models via Unsupervised Reward Optimization

Lanqing Li , Shentong Mo , Yang Yu , Pheng-Ann Heng This is my paper

Pith reviewed 2026-06-26 21:39 UTC · model grok-4.3

classification 💻 cs.LG
keywords protein language modelsunsupervised reward optimizationcontrollable generationoffline RLHFbiomolecular designself-improvementproxy rewards
0
0 comments X

The pith

Task-agnostic rewards from model uncertainty and semantic consistency enable unsupervised steering of protein language models without labels or experiments.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that protein language models can improve their controllability on compositional prompts by optimizing against proxy rewards computed from their own uncertainty and from semantic consistency signals in protein representation models. These proxies correlate with controllability across base models and temperatures, allowing two new offline algorithms, Soft Reward Optimization and Binarized Reward Optimization, to maximize a standard RLHF-style objective. Experiments show the resulting models outperform DPO and KTO baselines while approaching the performance of an oracle that uses ground-truth rewards. The approach yields higher pass@k coverage than the original models across sampling temperatures, model sizes, and protein families. This removes the need for curated preference data or wet-lab feedback in post-training adaptation.

Core claim

Task-agnostic rewards that combine intrinsic model uncertainty with extrinsic semantic consistency informed by protein representation models exhibit strong correlation with controllability measures, and two offline algorithms, Soft Reward Optimization (SRO) and Binarized Reward Optimization (BRO), maximize the classical RLHF objective induced by these proxies to produce steerable protein generation.

What carries the argument

Task-agnostic proxy rewards formed by combining intrinsic model uncertainty with extrinsic semantic consistency, optimized through the SRO and BRO offline algorithms.

If this is right

  • SRO and BRO both outperform DPO and KTO on compositional out-of-distribution prompts.
  • Performance approaches that of an oracle reward model across multiple sampling temperatures, model scales, and protein families.
  • Fine-tuned models achieve consistently higher coverage than the base model in pass@k evaluations.
  • Steerable biomolecular design becomes feasible in regimes where labeled preferences or experimental feedback are unavailable.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same uncertainty-plus-consistency proxy construction may transfer to other sequence-generation domains if the correlation with controllability holds there.
  • Iterative self-improvement loops become possible by repeatedly applying the method to the model's own outputs.
  • The framework could reduce dependence on wet-lab validation for initial steering of generative models in biomolecular design.

Load-bearing premise

The proxy rewards derived from uncertainty and semantic consistency accurately stand in for actual controllability and can drive RLHF-style optimization without ground-truth labels.

What would settle it

A controlled test in which models fine-tuned with SRO or BRO show no gain, or a loss, in measured controllability on held-out compositional prompts relative to the base model or to DPO/KTO.

Figures

Figures reproduced from arXiv: 2606.18961 by Lanqing Li, Pheng-Ann Heng, Shentong Mo, Yang Yu.

Figure 1
Figure 1. Figure 1: Correlations between intrinsic/extrinsic rewards and the ground-truth label in terms of AUROC on (a) Pfam700 prompt set for Func2Seq task, and DRAME prompt set for (b) Func2Seq and (c) Struct2Seq tasks, across five sampling temperatures. Promptwise AUROC evaluates the correlation within the generated sequences for each prompt. decreasing in temperature. In contrast, L1-mean always peaks at the critical tem… view at source ↗
Figure 2
Figure 2. Figure 2: Pass@k curves for Progen2-small-mix7 on Pfam700 Func2Seq tasks. For each prompt, N = 64 sequences are sampled for evaluation. 9 [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
read the original abstract

Protein language models (PLMs) have emerged as powerful tools for controllable biomolecular design, yet their post-training adaptation typically relies on costly wet-lab validation or curated preference datasets. To overcome this supervision bottleneck, we introduce unsupervised reward optimization of PLMs, a comprehensive framework for steerable protein generation without ground-truth labels. Our key insight is that task-agnostic rewards, which combine intrinsic model uncertainty with extrinsic semantic consistency informed by protein representation models, exhibit strong correlation with controllability measures across base models and temperature regimes. Building upon this discovery, we propose two offline algorithms: Soft Reward Optimization (SRO) and Binarized Reward Optimization (BRO), which effectively maximize the classical RLHF objective induced by these proxy rewards. Extensive experiments on compositional out-of-distribution prompts demonstrate that both methods significantly outperform competitive baselines (DPO, KTO), while approaching oracle performance across multiple sampling temperatures, model scales and protein families. Moreover, PLMs fine-tuned with unsupervised rewards can achieve consistently higher coverage compared to their base model in pass@k evaluations. By enabling self-improvement of PLMs through their own generated experience, our framework provides a scalable pathway toward controllable biomolecular design in settings where labeled preferences or experimental feedback are scarce or unavailable.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The manuscript introduces unsupervised reward optimization for protein language models (PLMs), claiming that task-agnostic proxy rewards—combining intrinsic model uncertainty with extrinsic semantic consistency from protein representation models—strongly correlate with controllability across base models and temperatures. It proposes two offline algorithms, Soft Reward Optimization (SRO) and Binarized Reward Optimization (BRO), to maximize the induced RLHF objective without ground-truth labels, and reports that these methods outperform DPO and KTO on compositional out-of-distribution prompts while approaching oracle performance and improving pass@k coverage across scales and families.

Significance. If the reported correlations and performance gains hold under the held-out controllability metrics, this provides a scalable, label-free pathway for self-improvement of PLMs in biomolecular design. The experiments across multiple temperatures, model scales, protein families, and comparisons to DPO/KTO supply independent checks that address potential circularity concerns in the proxy rewards, strengthening the case for practical utility where experimental feedback is unavailable.

minor comments (3)
  1. The abstract asserts strong correlation and outperformance but would benefit from one or two key quantitative metrics (e.g., correlation coefficient or relative improvement) to allow readers to assess the claims without the full text.
  2. [§3] §3 (method): the precise formulation of the combined reward (uncertainty + semantic consistency) and how it induces the classical RLHF objective should include an explicit equation for reproducibility.
  3. Figure captions for the correlation plots and pass@k results could explicitly state the number of independent runs and error bars to clarify statistical robustness.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our manuscript, the recognition of its potential impact, and the recommendation for minor revision. We are pleased that the experiments across scales, temperatures, and families were viewed as addressing potential concerns about circularity in the proxy rewards.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper frames the correlation between its task-agnostic proxy rewards (intrinsic uncertainty + extrinsic semantic consistency from separate protein representation models) and controllability measures as an empirical observation tested across base models, temperatures, and families. SRO and BRO are then defined as offline maximizers of the induced RLHF objective; the central results rest on held-out pass@k, compositional OOD, and baseline comparisons rather than any definitional reduction, fitted-input renaming, or self-citation chain. No equation or step equates a claimed prediction to its own construction inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that proxy rewards from uncertainty and semantic consistency are effective without external labels; no free parameters or invented entities are mentioned in the abstract.

axioms (1)
  • domain assumption Task-agnostic rewards combining model uncertainty and semantic consistency correlate with controllability measures.
    Stated as the key insight upon which the entire framework is built.

pith-pipeline@v0.9.1-grok · 5755 in / 1274 out tokens · 32604 ms · 2026-06-26T21:39:12.931839+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

92 extracted references · 23 canonical work pages · 8 internal anchors

  1. [1]

    Lawrence Zitnick, Jerry Ma, and Rob Fergus

    Alexander Rives, Joshua Meier, Tom Sercu, Siddharth Goyal, Zeming Lin, Jason Liu, Demi Guo, Myle Ott, C. Lawrence Zitnick, Jerry Ma, and Rob Fergus. Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences.PNAS, 2019. doi: 10.1101/622803. URLhttps://www.biorxiv.org/content/10.1101/622803v4

    work page doi:10.1101/622803 2019

  2. [2]

    Unified rational protein engineering with sequence-based deep representation learning

    Ethan C Alley, Grigory Khimulya, Surojit Biswas, Mohammed AlQuraishi, and George M Church. Unified rational protein engineering with sequence-based deep representation learning. Nature methods, 16(12):1315–1322, 2019

    2019

  3. [3]

    Prottrans: Toward understanding the language of life through self-supervised learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(10):7112–7127,

    Ahmed Elnaggar, Michael Heinzinger, Christian Dallago, Ghalia Rehawi, Yu Wang, Llion Jones, Tom Gibbs, Tamas Feher, Christoph Angerer, Martin Steinegger, Debsindhu Bhowmik, and Burkhard Rost. Prottrans: Toward understanding the language of life through self-supervised learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(10):7112–7127,

  4. [4]

    doi: 10.1109/TPAMI.2021.3095381

    work page doi:10.1109/tpami.2021.3095381 2021

  5. [5]

    Highly accurate protein structure prediction with alphafold.nature, 596(7873):583–589, 2021

    John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ron- neberger, Kathryn Tunyasuvunakool, Russ Bates, Augustin Žídek, Anna Potapenko, et al. Highly accurate protein structure prediction with alphafold.nature, 596(7873):583–589, 2021

    2021

  6. [6]

    Accurate prediction of protein structures and interactions using a three-track neural network.Science, 373(6557):871–876, 2021

    Minkyung Baek, Frank DiMaio, Ivan Anishchenko, Justas Dauparas, Sergey Ovchinnikov, Gyu Rie Lee, Jue Wang, Qian Cong, Lisa N Kinch, R Dustin Schaeffer, et al. Accurate prediction of protein structures and interactions using a three-track neural network.Science, 373(6557):871–876, 2021

    2021

  7. [7]

    Interpretable rna foundation model from unannotated data for highly accurate rna structure and function predictions.arXiv preprint arXiv:2204.00300, 2022

    Jiayang Chen, Zhihang Hu, Siqi Sun, Qingxiong Tan, Yixuan Wang, Qinze Yu, Licheng Zong, Liang Hong, Jin Xiao, Tao Shen, et al. Interpretable rna foundation model from unannotated data for highly accurate rna structure and function predictions.arXiv preprint arXiv:2204.00300, 2022

    work page arXiv 2022

  8. [8]

    Large language models generate functional protein sequences across diverse families.Nature biotechnology, 41 (8):1099–1106, 2023

    Ali Madani, Ben Krause, Eric R Greene, Subu Subramanian, Benjamin P Mohr, James M Holton, Jose Luis Olmos Jr, Caiming Xiong, Zachary Z Sun, Richard Socher, et al. Large language models generate functional protein sequences across diverse families.Nature biotechnology, 41 (8):1099–1106, 2023

    2023

  9. [9]

    De novo design of protein structure and function with rfdiffusion.Nature, 620(7976):1089–1100, 2023

    Joseph L Watson, David Juergens, Nathaniel R Bennett, Brian L Trippe, Jason Yim, Helen E Eisenach, Woody Ahern, Andrew J Borst, Robert J Ragotte, Lukas F Milles, et al. De novo design of protein structure and function with rfdiffusion.Nature, 620(7976):1089–1100, 2023

    2023

  10. [10]

    Saprot: Protein language modeling with structure-aware vocabulary

    Jin Su, Chenchen Han, Yuyang Zhou, Junjie Shan, Xibin Zhou, and Fajie Yuan. Saprot: Protein language modeling with structure-aware vocabulary. InThe Twelfth International Conference on Learning Representations, 2024

    2024

  11. [11]

    Sequence modeling and design from molecular to genome scale with evo.Science, 386(6723):eado9336, 2024

    Eric Nguyen, Michael Poli, Matthew G Durrant, Brian Kang, Dhruva Katrekar, David B Li, Liam J Bartie, Armin W Thomas, Samuel H King, Garyk Brixi, et al. Sequence modeling and design from molecular to genome scale with evo.Science, 386(6723):eado9336, 2024

    2024

  12. [12]

    Simulating 500 million years of evolution with a language model.Science, 387(6736):850–858, 2025

    Thomas Hayes, Roshan Rao, Halil Akin, Nicholas J Sofroniew, Deniz Oktay, Zeming Lin, Robert Verkuil, Vincent Q Tran, Jonathan Deaton, Marius Wiggert, et al. Simulating 500 million years of evolution with a language model.Science, 387(6736):850–858, 2025

    2025

  13. [13]

    Scalable emulation of protein equilibrium ensembles with generative deep learning.Science, 389(6761): eadv9817, 2025

    Sarah Lewis, Tim Hempel, José Jiménez-Luna, Michael Gastegger, Yu Xie, Andrew YK Foong, Victor García Satorras, Osama Abdin, Bastiaan S Veeling, Iryna Zaporozhets, et al. Scalable emulation of protein equilibrium ensembles with generative deep learning.Science, 389(6761): eadv9817, 2025

    2025

  14. [14]

    Boltz-2: Towards accurate and efficient binding affinity prediction.BioRxiv, 2025

    Saro Passaro, Gabriele Corso, Jeremy Wohlwend, Mateo Reveiz, Stephan Thaler, Vignesh Ram Somnath, Noah Getz, Tally Portnoi, Julien Roy, Hannes Stark, et al. Boltz-2: Towards accurate and efficient binding affinity prediction.BioRxiv, 2025

    2025

  15. [15]

    xtrimopglm: unified 100-billion-parameter pretrained transformer for deciphering the language of proteins.Nature Methods, 22(5):1028–1039, 2025

    Bo Chen, Xingyi Cheng, Pan Li, Yangli-ao Geng, Jing Gong, Shen Li, Zhilei Bei, Xu Tan, Boyan Wang, Xin Zeng, et al. xtrimopglm: unified 100-billion-parameter pretrained transformer for deciphering the language of proteins.Nature Methods, 22(5):1028–1039, 2025. 10

    2025

  16. [16]

    Sadit: Efficient protein backbone design via latent structural tokenization and diffusion transformers.arXiv preprint arXiv:2602.06706, 2026

    Shentong Mo and Lanqing Li. Sadit: Efficient protein backbone design via latent structural tokenization and diffusion transformers.arXiv preprint arXiv:2602.06706, 2026

    work page arXiv 2026

  17. [17]

    Modeling aspects of the language of life through transfer-learning protein sequences.BMC bioinformatics, 20(1):723, 2019

    Michael Heinzinger, Ahmed Elnaggar, Yu Wang, Christian Dallago, Dmitrii Nechaev, Florian Matthes, and Burkhard Rost. Modeling aspects of the language of life through transfer-learning protein sequences.BMC bioinformatics, 20(1):723, 2019

    2019

  18. [18]

    Language models enable zero-shot prediction of the effects of mutations on protein function.Advances in neural information processing systems, 34:29287–29303, 2021

    Joshua Meier, Roshan Rao, Robert Verkuil, Jason Liu, Tom Sercu, and Alex Rives. Language models enable zero-shot prediction of the effects of mutations on protein function.Advances in neural information processing systems, 34:29287–29303, 2021

    2021

  19. [19]

    Transformer protein language models are unsupervised structure learners

    Roshan Rao, Joshua Meier, Tom Sercu, Sergey Ovchinnikov, and Alexander Rives. Transformer protein language models are unsupervised structure learners. InInternational Conference on Learning Representations, 2021

    2021

  20. [20]

    Progen2: exploring the boundaries of protein language models.Cell systems, 14(11):968–978, 2023

    Erik Nijkamp, Jeffrey A Ruffolo, Eli N Weinstein, Nikhil Naik, and Ali Madani. Progen2: exploring the boundaries of protein language models.Cell systems, 14(11):968–978, 2023

    2023

  21. [21]

    Controllable protein design with language models.Nature Machine Intelligence, 4(6):521–532, 2022

    Noelia Ferruz and Birte Höcker. Controllable protein design with language models.Nature Machine Intelligence, 4(6):521–532, 2022

    2022

  22. [22]

    Aligning protein generative models with experimental fitness via direct preference optimization.bioRxiv, pages 2024–05, 2024

    Talal Widatalla, Rafael Rafailov, and Brian Hie. Aligning protein generative models with experimental fitness via direct preference optimization.bioRxiv, pages 2024–05, 2024

    2024

  23. [23]

    Guiding generative protein language models with reinforcement learning, 2024

    Filippo Stocco, Maria Artigues-Lleixa, Andrea Hunklinger, Talal Widatalla, Marc Guell, and Noelia Ferruz. Guiding generative protein language models with reinforcement learning, 2024. URLhttps://arxiv.org/abs/2412.12979

    work page arXiv 2024

  24. [24]

    Fine-Tuning Language Models from Human Preferences

    Daniel M Ziegler, Nisan Stiennon, Jeffrey Wu, Tom B Brown, Alec Radford, Dario Amodei, Paul Christiano, and Geoffrey Irving. Fine-tuning language models from human preferences. arXiv preprint arXiv:1909.08593, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1909

  25. [25]

    Learning to summarize with human feedback

    Nisan Stiennon, Long Ouyang, Jeffrey Wu, Daniel Ziegler, Ryan Lowe, Chelsea V oss, Alec Radford, Dario Amodei, and Paul F Christiano. Learning to summarize with human feedback. Advances in neural information processing systems, 33:3008–3021, 2020

    2020

  26. [26]

    Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback

    Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, et al. Training a helpful and harmless assistant with reinforcement learning from human feedback.arXiv preprint arXiv:2204.05862, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  27. [27]

    Training language models to follow instructions with human feedback.Advances in neural information processing systems, 35:27730–27744, 2022

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback.Advances in neural information processing systems, 35:27730–27744, 2022

    2022

  28. [28]

    Pretraining language models with human preferences

    Tomasz Korbak, Kejian Shi, Angelica Chen, Rasika Vinayak Bhalerao, Christopher Buckley, Jason Phang, Samuel R Bowman, and Ethan Perez. Pretraining language models with human preferences. InInternational conference on machine learning, pages 17506–17533. PMLR, 2023

    2023

  29. [29]

    DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  30. [30]

    Deepseek-r1 incentivizes reasoning in llms through reinforcement learning.Nature, 645(8081):633–638, 2025

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Peiyi Wang, Qihao Zhu, Runxin Xu, Ruoyu Zhang, Shirong Ma, Xiao Bi, et al. Deepseek-r1 incentivizes reasoning in llms through reinforcement learning.Nature, 645(8081):633–638, 2025

    2025

  31. [31]

    Qwq-32b: Embracing the power of reinforcement learning, 2025

    Qwen Team. Qwq-32b: Embracing the power of reinforcement learning, 2025. URL https: //qwenlm.github.io/blog/qwq-32b/

    2025

  32. [32]

    Qwen3 Technical Report

    An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al. Qwen3 technical report.arXiv preprint arXiv:2505.09388, 2025. 11

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  33. [33]

    Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities

    Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, Marcel Blistein, Ori Ram, Dan Zhang, Evan Rosen, et al. Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and next generation agentic capabilities.arXiv preprint arXiv:2507.06261, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  34. [34]

    Weak-to-strong generalization: Eliciting strong capabilities with weak supervision

    Collin Burns, Pavel Izmailov, Jan Hendrik Kirchner, Bowen Baker, Leo Gao, Leopold As- chenbrenner, Yining Chen, Adrien Ecoffet, Manas Joglekar, Jan Leike, et al. Weak-to-strong generalization: Eliciting strong capabilities with weak supervision. InForty-first International Conference on Machine Learning

  35. [35]

    Welcome to the era of experience.Google AI, 1:11, 2025

    David Silver and Richard S Sutton. Welcome to the era of experience.Google AI, 1:11, 2025

    2025

  36. [36]

    How far can unsupervised RLVR scale LLM training? InThe Thirteenth International Conference on Learning Representations, 2026

    Bingxiang He, Yuxin Zuo, Zeyuan Liu, Shangziqi Zhao, Zixuan Fu, Junlin Yang, Cheng Qian, Kaiyan Zhang, Yuchen Fan, Ganqu Cui, Xiusi Chen, Youbang Sun, Xingtai Lv, Xuekai Zhu, Li Sheng, Ran Li, Huan ang Gao, Yuchen Zhang, Lifan Yuan, Bowen Zhou, et al. How far can unsupervised RLVR scale LLM training? InThe Thirteenth International Conference on Learning R...

    2026

  37. [37]

    Generalist biological artificial intelligence in modeling the language of life.Nature Biotechnology, pages 1–16, 2026

    Vishwanatha M Rao, Serena Zhang, Brian S Plosky, Patrick D Hsu, Bo Wang, James Zou, Marinka Zitnik, Eric J Topol, and Pranav Rajpurkar. Generalist biological artificial intelligence in modeling the language of life.Nature Biotechnology, pages 1–16, 2026

    2026

  38. [38]

    Absolute zero: Reinforced self-play reasoning with zero data

    Andrew Zhao, Yiran Wu, Yang Yue, Tong Wu, Quentin Xu, Matthieu Lin, Shenzhi Wang, Qingyun Wu, Zilong Zheng, and Gao Huang. Absolute zero: Reinforced self-play reasoning with zero data. InThe Thirty-ninth Annual Conference on Neural Information Processing Systems

  39. [39]

    TTRL: Test-time reinforcement learning

    Yuxin Zuo, Kaiyan Zhang, Li Sheng, Shang Qu, Ganqu Cui, Xuekai Zhu, Haozhan Li, Yuchen Zhang, Xinwei Long, Ermo Hua, Biqing Qi, Youbang Sun, Zhiyuan Ma, Lifan Yuan, Ning Ding, and Bowen Zhou. TTRL: Test-time reinforcement learning. InThirty-ninth Conference on Neural Information Processing Systems, 2025. URL https://openreview.net/forum? id=VuVhgEiu20. Po...

    2025

  40. [40]

    Right question is already half the answer: Fully unsupervised llm reasoning incentivization

    Qingyang Zhang, Haitao Wu, Changqing Zhang, Peilin Zhao, and Yatao Bian. Right question is already half the answer: Fully unsupervised llm reasoning incentivization. InThirty-ninth Conference on Neural Information Processing Systems, 2025. URL https://openreview. net/forum?id=k8Mim6RI5O. Spotlight Presentation

    2025

  41. [41]

    Direct preference optimization: Your language model is secretly a reward model

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. Direct preference optimization: Your language model is secretly a reward model. Advances in neural information processing systems, 36:53728–53741, 2023

    2023

  42. [42]

    Model alignment as prospect theoretic optimization

    Kawin Ethayarajh, Winnie Xu, Niklas Muennighoff, Dan Jurafsky, and Douwe Kiela. Model alignment as prospect theoretic optimization. InInternational Conference on Machine Learning, pages 12634–12651. PMLR, 2024

    2024

  43. [43]

    Pfam: the protein families database.Nucleic acids research, 42(D1):D222–D230, 2014

    Robert D Finn, Alex Bateman, Jody Clements, Penelope Coggill, Ruth Y Eberhardt, Sean R Eddy, Andreas Heger, Kirstie Hetherington, Liisa Holm, Jaina Mistry, et al. Pfam: the protein families database.Nucleic acids research, 42(D1):D222–D230, 2014

    2014

  44. [44]

    Generative artificial intelligence for de novo protein design.Current Opinion in Structural Biology, 86:102794, 2024

    Adam Winnifrith, Carlos Outeiral, and Brian L Hie. Generative artificial intelligence for de novo protein design.Current Opinion in Structural Biology, 86:102794, 2024

    2024

  45. [45]

    De novo protein design—from new structures to programmable functions

    Tanja Kortemme. De novo protein design—from new structures to programmable functions. Cell, 187(3):526–544, 2024

    2024

  46. [46]

    The past, present and future of de novo protein design.Nature, 652(8112):1139–1152, 2026

    Wei Yang, Shunzhi Wang, Gyu Rie Lee, Jason Z Zhang, Alexis Courbet, David Juergens, Xinru Wang, Thomas Schlichthaerle, Mohamad Abedi, Robert Ragotte, et al. The past, present and future of de novo protein design.Nature, 652(8112):1139–1152, 2026

    2026

  47. [47]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017. 12

    2017

  48. [48]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. InProceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers), pages 4171–4186, 2019

    2019

  49. [49]

    Prollama: A protein large language model for multi-task protein language processing.IEEE Transactions on Artificial Intelligence, 2025

    Liuzhenghao Lv, Zongying Lin, Hao Li, Yuyang Liu, Jiaxi Cui, Calvin Yu-Chian Chen, Li Yuan, and Yonghong Tian. Prollama: A protein large language model for multi-task protein language processing.IEEE Transactions on Artificial Intelligence, 2025

    2025

  50. [50]

    Rapid in silico directed evolution by a protein language model with evolvepro.Science, 387 (6732):eadr6006, 2024

    Kaiyi Jiang, Zhaoqing Yan, Matteo Di Bernardo, Samantha R Sgrizzi, Lukas Villiger, Alisan Kayabolen, BJ Kim, Josephine K Carscadden, Masahiro Hiraizumi, Hiroshi Nishimasu, et al. Rapid in silico directed evolution by a protein language model with evolvepro.Science, 387 (6732):eadr6006, 2024

    2024

  51. [51]

    Machine-learning-guided directed evolution for protein engineering.Nature methods, 16(8):687–694, 2019

    Kevin K Yang, Zachary Wu, and Frances H Arnold. Machine-learning-guided directed evolution for protein engineering.Nature methods, 16(8):687–694, 2019

    2019

  52. [52]

    Steering protein language models

    Long-Kai Huang, Rongyi Zhu, Bing He, and Jianhua Yao. Steering protein language models. InInternational Conference on Machine Learning, pages 26247–26260. PMLR, 2025

    2025

  53. [53]

    Evaluating Large Language Models in Scientific Discovery

    Zhangde Song, Jieyu Lu, Yuanqi Du, Botao Yu, Thomas M Pruyn, Yue Huang, Kehan Guo, Xiuzhe Luo, Yuanhao Qu, Yi Qu, et al. Evaluating large language models in scientific discovery. arXiv preprint arXiv:2512.15567, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  54. [54]

    Learning to reason without external rewards

    Xuandong Zhao, Zhewei Kang, Aosong Feng, Sergey Levine, and Dawn Song. Learning to reason without external rewards. 2026. URL https://openreview.net/forum?id= OU9nFEYR2M

    2026

  55. [55]

    The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models

    Ganqu Cui, Yuchen Zhang, Jiacheng Chen, Lifan Yuan, Zhi Wang, Yuxin Zuo, Haozhan Li, Yuchen Fan, Huayu Chen, Weize Chen, et al. The entropy mechanism of reinforcement learning for reasoning language models.arXiv preprint arXiv:2505.22617, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  56. [56]

    The unreasonable effectiveness of entropy minimization in LLM reasoning

    Shivam Agarwal, Zimin Zhang, Lifan Yuan, Jiawei Han, and Hao Peng. The unreasonable effectiveness of entropy minimization in LLM reasoning. InThirty-ninth Conference on Neural Information Processing Systems, 2025. URL https://openreview.net/forum?id= UfFTBEsLgI. Poster Presentation

    2025

  57. [57]

    Can large reasoning models self-train?arXiv preprint arXiv:2505.21444,

    Sheikh Shafayat, Fahim Tajwar, Ruslan Salakhutdinov, Jeff Schneider, and Andrea Zanette. Can large reasoning models self-train?arXiv preprint arXiv:2505.21444, 2025

    work page arXiv 2025

  58. [58]

    No free lunch: Rethinking internal feedback for llm reasoning.arXiv preprint arXiv:2506.17219, 2025

    Yanzhi Zhang, Zhaoxi Zhang, Haoxiang Guan, Yilin Cheng, Yitong Duan, Chen Wang, Yue Wang, Shuxin Zheng, and Jiyan He. No free lunch: Rethinking internal feedback for llm reasoning.arXiv preprint arXiv:2506.17219, 2025

    work page arXiv 2025

  59. [59]

    Reinforcement pre-training.arXiv preprint arXiv:2506.08007, 2025

    Qingxiu Dong, Li Dong, Yao Tang, Tianzhu Ye, Yutao Sun, Zhifang Sui, and Furu Wei. Reinforcement pre-training.arXiv preprint arXiv:2506.08007, 2025

    work page arXiv 2025

  60. [60]

    Rlp: Reinforcement as a pretraining objective.arXiv preprint arXiv:2510.01265, 2025

    Ali Hatamizadeh, Syeda Nahida Akter, Shrimai Prabhumoye, Jan Kautz, Mostofa Patwary, Mohammad Shoeybi, Bryan Catanzaro, and Yejin Choi. Rlp: Reinforcement as a pretraining objective.arXiv preprint arXiv:2510.01265, 2025

    work page arXiv 2025

  61. [61]

    Dupo: Enabling reliable llm self-verification via dual preference optimization.arXiv preprint arXiv:2508.14460, 2025

    Shuaijie She, Yu Bao, Yu Lu, Lu Xu, Tao Li, Wenhao Zhu, Shujian Huang, Shanbo Cheng, Lu Lu, and Yuxuan Wang. Dupo: Enabling reliable llm self-verification via dual preference optimization.arXiv preprint arXiv:2508.14460, 2025

    work page arXiv 2025

  62. [62]

    Nemotron- crossthink: Scaling self-learning beyond math reasoning

    Syeda Nahida Akter, Shrimai Prabhumoye, Matvei Novikov, Seungju Han, Ying Lin, Evelina Bakhturina, Eric Nyberg, Yejin Choi, Mostofa Patwary, Mohammad Shoeybi, et al. Nemotron- crossthink: Scaling self-learning beyond math reasoning. InProceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Lon...

    2026

  63. [63]

    Ladder: Self-improving llms through recursive problem decomposition.arXiv preprint arXiv:2503.00735, 2025

    Toby Simonds and Akira Yoshiyama. Ladder: Self-improving llms through recursive problem decomposition.arXiv preprint arXiv:2503.00735, 2025. 13

    work page arXiv 2025

  64. [64]

    Deepseekmath-v2: Towards self-verifiable mathematical reasoning

    Zhihong Shao, Yuxiang Luo, Chengda Lu, ZZ Ren, Jiewen Hu, Tian Ye, Zhibin Gou, Shirong Ma, and Xiaokang Zhang. Deepseekmath-v2: Towards self-verifiable mathematical reasoning. arXiv preprint arXiv:2511.22570, 2025

    work page arXiv 2025

  65. [65]

    Olympiad-level formal mathematical reasoning with reinforcement learning.Nature, pages 1–3, 2025

    Thomas Hubert, Rishi Mehta, Laurent Sartran, Miklós Z Horváth, Goran Žuži´c, Eric Wieser, Aja Huang, Julian Schrittwieser, Yannick Schroecker, Hussain Masoom, et al. Olympiad-level formal mathematical reasoning with reinforcement learning.Nature, pages 1–3, 2025

    2025

  66. [66]

    Esm cambrian: Revealing the mysteries of proteins with unsupervised learning,

    ESM Team. Esm cambrian: Revealing the mysteries of proteins with unsupervised learning,

  67. [67]

    URLhttps://evolutionaryscale.ai/blog/esm-cambrian

  68. [68]

    Finetuned language models are zero-shot learners

    Jason Wei, Maarten Bosma, Vincent Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. Finetuned language models are zero-shot learners. In International Conference on Learning Representations, 2022

    2022

  69. [69]

    Maximum entropy inverse reinforcement learning

    Brian D Ziebart, Andrew L Maas, J Andrew Bagnell, Anind K Dey, et al. Maximum entropy inverse reinforcement learning. InAaai, volume 8, pages 1433–1438. Chicago, IL, USA, 2008

    2008

  70. [70]

    Reinforcement learning with deep energy-based policies

    Tuomas Haarnoja, Haoran Tang, Pieter Abbeel, and Sergey Levine. Reinforcement learning with deep energy-based policies. InInternational conference on machine learning, pages 1352–1361. PMLR, 2017

    2017

  71. [71]

    Soft actor-critic: Off- policy maximum entropy deep reinforcement learning with a stochastic actor

    Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, and Sergey Levine. Soft actor-critic: Off- policy maximum entropy deep reinforcement learning with a stochastic actor. InInternational conference on machine learning, pages 1861–1870. Pmlr, 2018

    2018

  72. [72]

    Semantic uncertainty: Linguistic invariances for uncertainty estimation in natural language generation

    Lorenz Kuhn, Yarin Gal, and Sebastian Farquhar. Semantic uncertainty: Linguistic invariances for uncertainty estimation in natural language generation. InThe Eleventh International Conference on Learning Representations, 2023

    2023

  73. [73]

    Detecting hallucinations in large language models using semantic entropy.Nature, 630(8017):625–630, 2024

    Sebastian Farquhar, Jannik Kossen, Lorenz Kuhn, and Yarin Gal. Detecting hallucinations in large language models using semantic entropy.Nature, 630(8017):625–630, 2024

    2024

  74. [74]

    Evaluating Large Language Models Trained on Code

    Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde De Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, et al. Evaluating large language models trained on code.arXiv preprint arXiv:2107.03374, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  75. [75]

    Hot or cold? adaptive temperature sampling for code generation with large language models

    Yuqi Zhu, Jia Li, Ge Li, YunFei Zhao, Zhi Jin, and Hong Mei. Hot or cold? adaptive temperature sampling for code generation with large language models. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 437–445, 2024

    2024

  76. [76]

    Edt: Improving large language models’ generation by entropy-based dynamic temperature sampling.arXiv preprint arXiv:2403.14541, 2024

    Shimao Zhang, Yu Bao, and Shujian Huang. Edt: Improving large language models’ generation by entropy-based dynamic temperature sampling.arXiv preprint arXiv:2403.14541, 2024

    work page arXiv 2024

  77. [77]

    Rank analysis of incomplete block designs: I

    Ralph Allan Bradley and Milton E Terry. Rank analysis of incomplete block designs: I. the method of paired comparisons.Biometrika, 39(3/4):324–345, 1952

    1952

  78. [78]

    https://doi.org/10.1109/BIBM62325.2024.10821806

    Hugo Hrbá ˇn and David Hoksza. Protein family sequence generation through progen2 fine- tuning. In2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 7037–7039, 2024. doi: 10.1109/BIBM62325.2024.10821712

    work page doi:10.1109/bibm62325.2024.10821712 2024

  79. [79]

    Interpro in 2022.Nucleic acids research, 51(D1):D418–D427, 2023

    Typhaine Paysan-Lafosse, Matthias Blum, Sara Chuguransky, Tiago Grego, Beatriz Lázaro Pinto, Gustavo A Salazar, Maxwell L Bileschi, Peer Bork, Alan Bridge, Lucy Colwell, et al. Interpro in 2022.Nucleic acids research, 51(D1):D418–D427, 2023

    2022

  80. [80]

    Learning inverse folding from millions of predicted structures

    Chloe Hsu, Robert Verkuil, Jason Liu, Zeming Lin, Brian Hie, Tom Sercu, Adam Lerer, and Alexander Rives. Learning inverse folding from millions of predicted structures. InInternational conference on machine learning, pages 8946–8970. PMLR, 2022

    2022

Showing first 80 references.