pith. sign in

arxiv: 2411.14721 · v2 · submitted 2024-11-22 · 💻 cs.CL · cs.LG· q-bio.QM

MolReFlect: Towards In-Context Fine-grained Alignments between Molecules and Texts

Jiatong Li , Yunqing Liu , Wei Liu , Jingdi Le , Di Zhang , Wenqi Fan , Dongzhan Zhou , Yuqiang Li
show 1 more author
Qing Li
This is my paper

Pith reviewed 2026-05-23 16:43 UTC · model grok-4.3

classification 💻 cs.CL cs.LGq-bio.QM
keywords fine-grained alignmentmolecule-caption translationteacher-student frameworklarge language modelsSMILES substructuresmolecular understanding
0
0 comments X

The pith

MolReFlect uses a teacher LLM to generate and refine mappings between SMILES substructures and caption phrases, then teaches them to a student LLM.

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

The paper tackles the problem that LLMs treat entire molecules as single inputs when linking them to text captions, which limits accuracy and explainability. It defines fine-grained alignments as precise links between a molecule's sub-structures and the specific phrases that describe their properties. MolReFlect introduces a teacher-student setup in which the teacher LLM creates and improves these phrase-to-substructure mappings from captions and SMILES strings, then passes the explicit alignments to the student LLM for learning. Experiments show this produces state-of-the-art results on molecule-caption translation without any expert human annotations for the alignments.

Core claim

MolReFlect is a teacher-student framework where a teacher LLM first generates and refines mappings between caption phrases and SMILES substructures and then explicitly teaches these detailed alignments to a student LLM, enabling significantly better performance than prior baselines on the molecule-caption translation task.

What carries the argument

The teacher-student framework that automatically generates and teaches fine-grained alignments between caption phrases and SMILES substructures.

If this is right

  • LLMs reach state-of-the-art results on molecule-caption translation tasks.
  • Models gain the ability to reason about molecules at the substructure level with explicit textual links.
  • Explainability improves because the model can point to which parts of a molecule correspond to which parts of its description.
  • The need for costly expert-labeled fine-grained alignment data is removed.

Where Pith is reading between the lines

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

  • The same teacher-student pattern could be tested on other sequence-to-text tasks such as protein function descriptions or material property captions.
  • If the generated alignments prove stable across different teacher models, the method could be used to bootstrap large-scale alignment datasets for training smaller specialized models.
  • Explicit substructure teaching may reduce hallucinated chemical claims in LLM outputs by grounding each phrase to a verifiable molecular fragment.

Load-bearing premise

The teacher LLM can generate and refine accurate, useful mappings between caption phrases and SMILES substructures without expert human annotation or verification.

What would settle it

A review by domain experts that finds most of the automatically generated phrase-to-substructure mappings to be chemically incorrect or unhelpful would show the framework does not produce reliable alignments.

Figures

Figures reproduced from arXiv: 2411.14721 by Di Zhang, Dongzhan Zhou, Jiatong Li, Jingdi Le, Qing Li, Wei Liu, Wenqi Fan, Yunqing Liu, Yuqiang Li.

Figure 1
Figure 1. Figure 1: An illustration of the alignments between the molecular space and the language space. The sub [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Comparisons of four different fine-tuning paradigms, including (a) Naive Supervised Fine-tuning (naive [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The overall framework of MolReFlect. string, respectively. 4.2 In-Context Selective Reflection Despite the powerful capabilities of LLMs, they can still generate answers with hallucinations (Yao et al., 2023). Their knowledge of chemistry is also limited due to the absence of domain pre￾training on chemical corpora, which can introduce noises into the zero-shot alignments. To mitigate these potential noise… view at source ↗
Figure 4
Figure 4. Figure 4: Embedding distributions of molecules and captions. [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Cases of Fine-grained Alignments. We could observe that the molecule structure and characteristics have [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Cases of Customized Examples for the Cap2Mol task. We follow the customized examples in ( [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Cases for the Mol2Cap task [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Cases for the Cap2Mol task [PITH_FULL_IMAGE:figures/full_fig_p018_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Prompt templates for Zero-shot Alignment Extraction. [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Prompt templates for In-Context Selective Reflection. [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Prompt templates for MolReFlect (w/o Fine-grained Alignments). [PITH_FULL_IMAGE:figures/full_fig_p021_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Prompt templates for Chain-of-Thought In-Context Molecule Tuning (CoT-ICMT). [PITH_FULL_IMAGE:figures/full_fig_p022_12.png] view at source ↗
read the original abstract

Molecule discovery is a pivotal research field, impacting everything from medicine to materials. Recently, Large Language Models (LLMs) have been widely adopted in molecular understanding and generation, serving as a bridge between the molecular space and the natural language space, yet the alignment between molecules and their corresponding captions remains a significant challenge. Previous endeavors typically treat molecules as monolithic inputs, lacking an intermediate reasoning process and sacrificing explainability. In this work, we define fine-grained alignments as the precise correspondence between a molecule's sub-structures and the textual phrases that explain their properties. These alignments are crucial for LLMs to understand molecules in a more accurate and explainable manner. Normally, such fine-grained alignments require expert annotation, which is both costly and time-consuming. To allow LLMs to automatically label and learn the fine-grained alignments, we propose MolReFlect, a novel teacher-student framework, where a teacher LLM first generates and refines mappings between caption phrases and SMILES substructures and then explicitly teaches these detailed alignments to a student LLM. Experimental results demonstrate that MolReFlect enables LLMs to significantly outperform previous baselines, achieving the state-of-the-art performance in the molecule-caption translation task. Our codes are available via: https://github.com/phenixace/MolReFlect.

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

2 major / 1 minor

Summary. The paper introduces MolReFlect, a teacher-student framework in which a teacher LLM automatically generates and refines mappings between textual phrases in molecule captions and substructures in SMILES strings; these mappings are then used to explicitly teach fine-grained alignments to a student LLM. The central claim is that this procedure yields state-of-the-art performance on the molecule-caption translation task without requiring expert human annotation.

Significance. If the teacher-generated alignments prove accurate and causally responsible for the reported gains, the method could improve both performance and explainability in molecular LLMs. The absence of any validation of the teacher outputs, however, leaves the significance of the empirical result unclear.

major comments (2)
  1. [Abstract] Abstract: the claim of 'state-of-the-art performance' is asserted without any reported metrics, baselines, dataset sizes, or ablation results, preventing verification of the central empirical claim.
  2. [Method] Method section (teacher-student framework): the load-bearing assumption that the teacher LLM produces accurate phrase-to-SMILES-substructure mappings is unsupported by any human validation, inter-annotator agreement, or proxy metric on a held-out set; without this, performance deltas cannot be attributed to fine-grained alignment rather than prompt length, extra examples, or other confounds.
minor comments (1)
  1. The GitHub link is given but no statement confirms that the released code reproduces the exact experimental pipeline and teacher prompts used in the paper.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive comments. We agree that the abstract should be more self-contained with quantitative results and that explicit validation of the teacher-generated alignments is needed to strengthen causal attribution. We will make both changes in the revision.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'state-of-the-art performance' is asserted without any reported metrics, baselines, dataset sizes, or ablation results, preventing verification of the central empirical claim.

    Authors: We agree that the abstract as submitted does not contain the supporting numbers. In the revised version we will expand the abstract to report the primary metrics (BLEU-2/4, ROUGE-L, METEOR), the two evaluation datasets with their sizes, the main baselines, and a one-sentence summary of the ablation results that isolate the contribution of the fine-grained alignments. These numbers already appear in the experimental section; the revision will simply surface them in the abstract. revision: yes

  2. Referee: [Method] Method section (teacher-student framework): the load-bearing assumption that the teacher LLM produces accurate phrase-to-SMILES-substructure mappings is unsupported by any human validation, inter-annotator agreement, or proxy metric on a held-out set; without this, performance deltas cannot be attributed to fine-grained alignment rather than prompt length, extra examples, or other confounds.

    Authors: We acknowledge that the original submission provides no direct validation of the teacher mappings. To address this, the revision will include (1) a human evaluation on a random sample of 200 teacher-generated alignments with inter-annotator agreement statistics, and (2) an ablation that compares the full teacher-student pipeline against a control that supplies the same number of additional tokens but without the phrase-to-substructure mappings. These additions will allow readers to assess both the quality of the teacher outputs and the extent to which the observed gains are attributable to the fine-grained alignments rather than length or example count. revision: yes

Circularity Check

0 steps flagged

No circularity: framework is an independent procedural proposal

full rationale

The paper describes a teacher-student LLM framework for generating phrase-SMILES mappings and reports experimental SOTA results on translation. No equations, fitted parameters, predictions derived from inputs, uniqueness theorems, or self-citations appear in the provided text. The central claim rests on empirical outcomes and the external assumption that the teacher produces accurate mappings, which is not reduced to the paper's own definitions or prior outputs by construction. This matches the default expectation of a non-circular methodological contribution.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework depends on the unverified capability of an LLM to produce accurate substructure-phrase mappings; no free parameters, new physical entities, or additional axioms beyond standard LLM prompting assumptions are stated.

axioms (1)
  • domain assumption An LLM prompted as teacher can generate and refine accurate mappings between textual phrases and SMILES substructures
    This capability is required for the teacher stage to produce usable training signals for the student.

pith-pipeline@v0.9.0 · 5793 in / 1122 out tokens · 19340 ms · 2026-05-23T16:43:22.156971+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Speak-to-Structure: Evaluating LLMs in Open-domain Natural Language-Driven Molecule Generation

    cs.CL 2024-12 unverdicted novelty 7.0

    S^2-Bench is a new one-to-many benchmark for natural language-driven molecule generation with three tasks, and OpenMolIns is an instruction dataset enabling Llama3.1-8B to outperform GPT-4o and Claude-3.5 on it.

  2. Mol-Debate: Multi-Agent Debate Improves Structural Reasoning in Molecular Design

    cs.AI 2026-04 unverdicted novelty 6.0

    Mol-Debate applies multi-agent debate in an iterative loop with perspective orchestration to achieve state-of-the-art text-guided molecular design, scoring 59.82% exact match on ChEBI-20 and 50.52% weighted success on...

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages · cited by 2 Pith papers · 6 internal anchors

  1. [1]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

    work page

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

    work page

  3. [3]

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  4. [4]

    Pervin Basaran and Emilio Rodr \' guez-Cerezo. 2008. Plant molecular farming: opportunities and challenges. Critical reviews in biotechnology, 28(3):153--172

    work page 2008

  5. [5]

    Yanda Chen, Ruiqi Zhong, Sheng Zha, George Karypis, and He He. 2022. Meta-learning via language model in-context tuning. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 719--730

    work page 2022

  6. [6]

    Dimitrios Christofidellis, Giorgio Giannone, Jannis Born, Ole Winther, Teodoro Laino, and Matteo Manica. 2023. Unifying molecular and textual representations via multi-task language modelling. arXiv preprint arXiv:2301.12586

    work page arXiv 2023

  7. [7]

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  8. [8]

    Carl Edwards, Tuan Lai, Kevin Ros, Garrett Honke, Kyunghyun Cho, and Heng Ji. 2022. https://aclanthology.org/2022.emnlp-main.26 Translation between molecules and natural language . In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 375--413, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics

    work page 2022

  9. [9]

    Carl Edwards, ChengXiang Zhai, and Heng Ji. 2021. Text2mol: Cross-modal molecule retrieval with natural language queries. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 595--607

    work page 2021

  10. [10]

    Chemical language models have problems with chemistry: A case study on molecule captioning task

    Veronika Ganeeva, Kuzma Khrabrov, Artur Kadurin, Andrey Savchenko, and Elena Tutubalina. Chemical language models have problems with chemistry: A case study on molecule captioning task. In The Second Tiny Papers Track at ICLR 2024

    work page 2024

  11. [11]

    Christina M Grozinger and Stuart L Schreiber. 2002. Deacetylase enzymes: biological functions and the use of small-molecule inhibitors. Chemistry & biology, 9(1):3--16

    work page 2002

  12. [12]

    Akon Higuchi, Tzu-Cheng Sung, Ting Wang, Qing-Dong Ling, S Suresh Kumar, Shih-Tien Hsu, and Akihiro Umezawa. 2023. Material design for next-generation mrna vaccines using lipid nanoparticles. Polymer Reviews, 63(2):394--436

    work page 2023

  13. [13]

    Edward J Hu, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. 2021. Lora: Low-rank adaptation of large language models. In International Conference on Learning Representations

    work page 2021

  14. [14]

    Ross Irwin, Spyridon Dimitriadis, Jiazhen He, and Esben Jannik Bjerrum. 2022. Chemformer: a pre-trained transformer for computational chemistry. Machine Learning: Science and Technology, 3(1):015022

    work page 2022

  15. [15]

    Albert Q Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, et al. 2023. Mistral 7b. arXiv preprint arXiv:2310.06825

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  16. [16]

    Michael J Keiser, John J Irwin, and Brian K Shoichet. 2010. The chemical basis of pharmacology. Biochemistry, 49(48):10267--10276

    work page 2010

  17. [17]

    Leszek Konieczny, Irena Roterman-Konieczna, and Pawe Sp \'o lnik. 2023. The structure and function of living organisms. In Systems Biology: Functional Strategies of Living Organisms, pages 1--52. Springer

    work page 2023

  18. [18]

    Mario Krenn, Florian H \"a se, AkshatKumar Nigam, Pascal Friederich, and Alan Aspuru-Guzik. 2020. Self-referencing embedded strings (selfies): A 100\ Machine Learning: Science and Technology, 1(4):045024

    work page 2020

  19. [19]

    Greg Landrum. 2013. Rdkit documentation. Release, 1(1-79):4

    work page 2013

  20. [20]

    Jiatong Li, Wei Liu, Zhihao Ding, Wenqi Fan, Yuqiang Li, and Qing Li. 2024 a . Large language models are in-context molecule learners. arXiv preprint arXiv:2403.04197

    work page arXiv 2024

  21. [21]

    Jiatong Li, Yunqing Liu, Wenqi Fan, Xiao-Yong Wei, Hui Liu, Jiliang Tang, and Qing Li. 2023 a . Empowering molecule discovery for molecule-caption translation with large language models: A chatgpt perspective. arXiv preprint arXiv:2306.06615

    work page arXiv 2023

  22. [22]

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. 2023 b . Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. arXiv preprint arXiv:2301.12597

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  23. [23]

    Ming Li, Lichang Chen, Jiuhai Chen, Shwai He, Jiuxiang Gu, and Tianyi Zhou. 2024 b . Selective reflection-tuning: Student-selected data recycling for llm instruction-tuning. arXiv preprint arXiv:2402.10110

    work page arXiv 2024

  24. [24]

    Sihang Li, Zhiyuan Liu, Yanchen Luo, Xiang Wang, Xiangnan He, Kenji Kawaguchi, Tat-Seng Chua, and Qi Tian. 2024 c . Towards 3d molecule-text interpretation in language models. arXiv preprint arXiv:2401.13923

    work page arXiv 2024

  25. [25]

    Zhiyuan Liu, Sihang Li, Yanchen Luo, Hao Fei, Yixin Cao, Kenji Kawaguchi, Xiang Wang, and Tat-Seng Chua. 2023. Molca: Molecular graph-language modeling with cross-modal projector and uni-modal adapter. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 15623--15638

    work page 2023

  26. [26]

    Van-Thuan Nguyen, Young Seop Kwon, and Man Bock Gu. 2017. Aptamer-based environmental biosensors for small molecule contaminants. Current opinion in biotechnology, 45:15--23

    work page 2017

  27. [27]

    Qizhi Pei, Wei Zhang, Jinhua Zhu, Kehan Wu, Kaiyuan Gao, Lijun Wu, Yingce Xia, and Rui Yan. 2023. Biot5: Enriching cross-modal integration in biology with chemical knowledge and natural language associations. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 1102--1123

    work page 2023

  28. [28]

    Chen Qian, Huayi Tang, Zhirui Yang, Hong Liang, and Yong Liu. 2023. Can large language models empower molecular property prediction? arXiv preprint arXiv:2307.07443

    work page arXiv 2023

  29. [29]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learning, pages 8748--8763. PMLR

    work page 2021

  30. [30]

    Fran c isco M Raymo and Silvia Giordani. 2001. Signal processing at the molecular level. Journal of the American Chemical Society, 123(19):4651--4652

    work page 2001

  31. [31]

    Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3982--3992

    work page 2019

  32. [32]

    Stephen Robertson, Hugo Zaragoza, et al. 2009. The probabilistic relevance framework: Bm25 and beyond. Foundations and Trends in Information Retrieval , 3(4):333--389

    work page 2009

  33. [33]

    Bing Su, Dazhao Du, Zhao Yang, Yujie Zhou, Jiangmeng Li, Anyi Rao, Hao Sun, Zhiwu Lu, and Ji-Rong Wen. 2022. A molecular multimodal foundation model associating molecule graphs with natural language. arXiv preprint arXiv:2209.05481

    work page arXiv 2022

  34. [34]

    Ross Taylor, Marcin Kardas, Guillem Cucurull, Thomas Scialom, Anthony Hartshorn, Elvis Saravia, Andrew Poulton, Viktor Kerkez, and Robert Stojnic. 2022. Galactica: A large language model for science. arXiv preprint arXiv:2211.09085

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  35. [35]

    Richard M Twyman, Eva Stoger, Stefan Schillberg, Paul Christou, and Rainer Fischer. 2003. Molecular farming in plants: host systems and expression technology. TRENDS in Biotechnology, 21(12):570--578

    work page 2003

  36. [36]

    Athanasios Valavanidis, Thomais Vlahogianni, Manos Dassenakis, and Michael Scoullos. 2006. Molecular biomarkers of oxidative stress in aquatic organisms in relation to toxic environmental pollutants. Ecotoxicology and environmental safety, 64(2):178--189

    work page 2006

  37. [37]

    Jason Wei, Maarten Bosma, Vincent Y Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. 2021. Finetuned language models are zero-shot learners. arXiv preprint arXiv:2109.01652

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  38. [38]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824--24837

    work page 2022

  39. [39]

    David Weininger. 1988. Smiles, a chemical language and information system. 1. introduction to methodology and encoding rules. Journal of chemical information and computer sciences, 28(1):31--36

    work page 1988

  40. [40]

    Zhenqin Wu, Bharath Ramsundar, Evan N Feinberg, Joseph Gomes, Caleb Geniesse, Aneesh S Pappu, Karl Leswing, and Vijay Pande. 2018. Moleculenet: a benchmark for molecular machine learning. Chemical science, 9(2):513--530

    work page 2018

  41. [41]

    Jun Xia, Chengshuai Zhao, Bozhen Hu, Zhangyang Gao, Cheng Tan, Yue Liu, Siyuan Li, and Stan Z Li. 2022. Mole-bert: Rethinking pre-training graph neural networks for molecules. In The Eleventh International Conference on Learning Representations

    work page 2022

  42. [42]

    Jia-Yu Yao, Kun-Peng Ning, Zhen-Hui Liu, Mu-Nan Ning, and Li Yuan. 2023. Llm lies: Hallucinations are not bugs, but features as adversarial examples. arXiv preprint arXiv:2310.01469

    work page arXiv 2023

  43. [43]

    Geyan Ye, Xibao Cai, Houtim Lai, Xing Wang, Junhong Huang, Longyue Wang, Wei Liu, and Xiangxiang Zeng. 2023. Drugassist: A large language model for molecule optimization. arXiv preprint arXiv:2401.10334

    work page arXiv 2023

  44. [44]

    Di Zhang, Wei Liu, Qian Tan, Jingdan Chen, Hang Yan, Yuliang Yan, Jiatong Li, Weiran Huang, Xiangyu Yue, Dongzhan Zhou, et al. 2024. Chemllm: A chemical large language model. arXiv preprint arXiv:2402.06852

    work page arXiv 2024