Pith. sign in

REVIEW 5 major objections 4 minor 1 cited by

Improving Linguistic Diversity of Large Language Models with Possibility Exploration Fine-Tuning

T0 review · 5 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Fine-tuning with possibility numbers makes one prompt yield many diverse replies.

desk verdict Useful diversity fine-tuning recipe, but the 'controllable possibility number' claim is untested and likely overreaches. read the letter →

arxiv 2412.03343 v1 pith:EL73KOFO submitted 2024-12-04 cs.CL cs.AI

Long Mai , Julie Carson-Berndsen This is my paper · ORCID
classification cs.CLcs.AI
keywords linguisticdiversitypossibilityexplorationfine-tuningone-to-manydataunlikelihoodtrainingsemanticdemographicbiasdialoguegenerationstory
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that the low linguistic diversity of LLM outputs can be fixed without added latency or compute by fine-tuning on one-to-many data where each training response is tagged with a randomly assigned 'possibility number.' The central idea is that asking the model to produce the response for possibility k during training, and using unlikelihood loss to push the same response away from other k values, teaches the model to associate different k's with semantically different response modes. On open-domain dialogue and story generation, the authors report that their method, PEFT, beats the base model and existing diversity techniques on semantic diversity while keeping coherence and inference speed at base-model levels. They also report that the increased semantic diversity carries over to the personas a chatbot invents, raising the demographic entropy of generated personas and thereby reducing representational bias.

What carries the argument

The PE (Possibility Exploration) dataset paired with a likelihood-unlikelihood training objective. Each training prompt is replicated $N$ times with $N$ semantically distinct responses, each assigned a possibility number $k$ sampled from $[1, M]$; the unlikelihood term then treats each response as a negative example for every other $k$ in the batch. This creates a one-to-one association between a possibility number and a response meaning, which is what lets the model generate a controllable set of diverse outputs at inference.

What would settle it

Generate responses for several different prompts with the same possibility number $k$ and different $k$ values, then measure the SBERT similarity of responses sharing the same $k$ across prompts. If same-$k$ responses are no more similar to each other than responses from different $k$'s, the controllable possibility-number mapping has not been learned.

Watch

Extended reading notes

Core claim

PEFT (Possibility Exploration Fine-Tuning) is a fine-tuning framework that treats open-ended generation as a one-to-many problem: for each prompt, the training set contains several distinct valid responses, each paired with a possibility number $k$ randomly drawn from $[1, M]$ with $M > N$. The model is trained with a maximum-likelihood term that teaches it to produce response $r_i$ when prompted with possibility number $k_i$, and an unlikelihood term that penalizes producing the same $r_i$ when a different possibility number is given, with penalties applied at the first token of each word. During inference, the user samples $L$ possibility numbers and generates the corresponding responses independently and simultaneously, so the number of distinct responses is controlled by the choice of $k$ and the cost is the same as generating $L$ responses with the base model. The paper's experiments on Mistral 7B and LLAMA 2 show that this increases semantic diversity (measured by $1 -$ SBERT pairwise similarity) while holding the percentage of incoherent responses low, and that the diversity gain transfers to persona generation, where PEFT raises the Shannon entropy of gender, age, location, education, and occupation attributes in chatbot-generated personas.

Load-bearing premise

The model must learn, from just four training examples per prompt, a reliable mapping from a randomly assigned possibility number to a distinct response meaning, and the paper never verifies that this mapping holds for prompts outside the training set.

Editorial extensions

If this is right

  • Fine-tuned with PEFT, a dialogue model can offer several semantically different replies per turn at base-model latency, which suits real-time chatbots.
  • PEFT combines with existing decoding methods such as temperature sampling and diverse beam search to push diversity further without architectural changes.
  • Because PEFT emphasizes semantic over lexical diversity, it also flattens demographic skew in persona generation, providing a fine-tuning-based route to debiasing.
  • The framework is task-agnostic: applying it to story generation reproduces the diversity and coherence gains seen in dialogue.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the controllability mapping from $k$ to semantics is reliable, PEFT could serve as a cheap way to sample diverse candidates for self-consistency or tree-of-thought reasoning, where candidate diversity is the bottleneck; the paper does not test this.
  • A direct extension would be to apply PEFT during the instruction-tuning phase rather than post-hoc, as the authors note in their limitations, which would make the diversity benefit available zero-shot for arbitrary tasks.
  • The same possibility-number conditioning could also act as a controllable diversity knob at inference time, but only if the mapping generalizes across prompts; measuring same-$k$ semantic consistency across different prompts would settle that.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 4 minor

Summary. This paper proposes Possibility Exploration Fine-Tuning (PEFT), a fine-tuning framework intended to increase the semantic diversity of LLM responses while preserving coherence and latency. The method constructs a one-to-many training set by using GPT-4o and List Prompting to generate four distinct responses per prompt, then augments each response with a randomly assigned possibility number k drawn from 1..M (with M=9). Training combines maximum likelihood on positive (prompt, k, response) triples with an unlikelihood loss whose negative samples reuse the same response under a different k, thereby penalizing the model for generating that response under the wrong possibility number. Experiments on open-domain dialogue (Mistral 7B and LLAMA 2 7B) and story generation report that PEFT improves lexical and semantic diversity scores, achieves a good diversity-coherence-latency trade-off compared to list prompting and a conditional variational framework, and increases the entropy of persona attributes, which the authors interpret as reduced demographic bias.

Significance. If the results hold, PEFT offers a practical, latency-neutral way to improve output diversity, and the demonstration that fine-tuning on one-to-many data can reduce demographic skew is of interest to the dialogue and fairness communities. The method is transparent and the repository is provided, which supports reproducibility. However, the paper's headline claim of a 'controllable possibility number' is not validated by any direct experiment, and the evaluation pipeline contains multiple circular uses of GPT-4o (training-data generation, coherence judging, persona extraction, and persona seeker). The absence of significance tests, error bars, or full temperature sweeps further weakens the quantitative claims. The core idea is defensible, but the current evidence does not support the strong wording in the abstract.

major comments (5)
  1. [Abstract and §4.2–4.3] The claim that PEFT responses correspond to a 'controllable possibility number' is not supported by the training design or the experiments. In §4.2, k_i is 'randomly sampled from [1,..,M]' for each response, so there is no cross-prompt semantic meaning attached to a particular k. The unlikelihood objective in §4.3 only forces a one-to-one mapping within each training batch (N=4 examples). The evaluation in §5 samples possibility numbers and measures set-level diversity, but never tests whether a fixed k produces semantically consistent responses across prompts or whether different k values reliably select distinct, predictable semantic modes. Without such a test, the abstract's central claim remains speculative; the method may simply be OTMFT plus a random seed. Please add a direct controllability experiment or explicitly demote this claim.
  2. [§5.4, Tables 1–3] The comparison reports only 'optimal temperatures for each method,' with the optimal temperature chosen on the same test set used for evaluation. This selection procedure can overfit the test set and makes the claimed improvements (e.g., PEFT 0.585 vs. OTMFT 0.522 in Table 1) hard to interpret. In addition, the test set contains only 300 prompts, and no significance tests, confidence intervals, or multiple-run variance are reported. The paper uses the word 'significantly' throughout, but there is no statistical support. Please report results for all temperatures, add bootstrap confidence intervals or repeated-seed variance, and, if possible, a significance test.
  3. [§5.2.2, §6, Appendix A.4] The coherence metric uses GPT-4o and LLAMA 3 with a fixed threshold (rating <6), yet the paper also uses GPT-4o to generate the training data, to serve as the persona seeker, and to extract persona attributes. This creates a circular evaluation pipeline: the fine-tuned model is trained to imitate GPT-4o's diverse responses and is then judged by the same model family. The bias-reduction result (Table 4) is especially vulnerable because the persona attributes are extracted by GPT-4o and the conversations are driven by ChatGPT. Please provide a human evaluation or at least an evaluation using a model that was not involved in data generation, and show the sensitivity of the incoherence measure to the threshold choice.
  4. [§4.3 and §6] The contribution of the unlikelihood term relative to the one-to-many data itself is not isolated. OTMFT already improves diversity over OTOFT, and the tables compare PEFT to OTMFT at different temperatures; the only head-to-head at the same temperature (t=0.75, Table 1: OTMFT 0.565 vs. PEFT 0.585, and t=1.0, Table 2: 0.579 vs. 0.583) shows small differences. An ablation that trains PEFT without the unlikelihood loss (i.e., pure OTMFT with random k) at identical settings would clarify whether the possibility-number conditioning adds value beyond the richer training data. Without this, the improvement over OTMFT may come only from the negative training, which could also be achieved by simpler data augmentation.
  5. [§5.1 and §4.1] The training data are generated by GPT-4o with List Prompting, so the diversity gains reported may reflect the diversity of the teacher outputs rather than the PEFT framework itself. The paper does not compare against fine-tuning on a larger one-to-many dataset generated by a different procedure, nor does it vary N (number of responses per prompt) or M (max possibility number). A controlled experiment isolating the effect of the data source and the number of responses would strengthen the generalizability claim. As it stands, the reader cannot tell whether PEFT would add diversity beyond simply training on any diverse set of GPT-4o outputs.
minor comments (4)
  1. [§4.3, Eq. (1)] The MLE loss sums over t starting at 0, but language-modeling likelihoods typically start at t=1 since r_0 is a start token. Please fix the index or clarify the tokenization convention.
  2. [§4.2 and Appendix A.3.2] The prompt template in A.3.2 uses '$k' in LaTeX-style math; this is likely a typo for the actual integer value. Please make the notation consistent.
  3. [General] The acronym PEFT is widely used for Parameter-Efficient Fine-Tuning; the paper's use of the same acronym for Possibility Exploration Fine-Tuning may confuse readers. Consider renaming or explicitly distinguishing it from the established term.
  4. [Table 3 and Table 5] The 'Incoh' column label is not defined in the captions; please spell out 'incoherence rate' in the table captions.

Circularity Check

0 steps flagged · score 0.0 of 10

No constructional circularity: the diversity and bias results are held-out empirical comparisons, and the GPT-4o evaluation overlap is an evaluation-validity concern rather than a derivation loop.

full rationale

The paper's central empirical claims—PEFT increases semantic diversity and improves bias entropy—are supported by held-out comparisons (300 dialogue contexts and 300 story-cloze test items) against the base model and baselines. The diversity metric (1 minus SBERT pairwise similarity) is external to the GPT-4o-generated training data, and the reported gains over OTMFT come from an additional unlikelihood objective plus random possibility-number conditioning; these are not fitted to the test set, so no prediction reduces to an input by construction. There are no self-citations, no imported uniqueness theorems, and no ansatz smuggled in via the authors' own prior work. The main validity concerns lie outside the circularity definition: the evaluation pipeline overlaps with the data-generation pipeline (GPT-4o generates training responses, judges coherence, extracts persona attributes, and ChatGPT drives persona conversations), which can bias absolute scores but does not make any claimed result equivalent to its inputs; and the "controllable possibility number" semantics is untested, since k is randomly assigned in Section 4.2 and no cross-prompt consistency check is reported. These are unvalidated assumptions or evaluation confounds, not circular constructions, so the circularity score remains 0.

Assumptions & free parameters 5 free parameters · 5 assumptions · 1 invented entities

The central claim rests on several hand-chosen hyperparameters (M, alpha, N, beta, per-method temperatures), on the assumption that GPT-4o is a valid teacher for diversity, and on the assumption that LLM-based metrics for coherence and persona attributes are valid. The invented possibility number is a new conditioning feature whose controllability is not independently validated.

free parameters (5)
  • M (max possibility number) = 9
    Hand-chosen upper bound for the random possibility number; no sensitivity analysis is provided.
  • alpha (unlikelihood weight) = 0.5
    Hand-chosen mixture weight for the unlikelihood loss; no ablation is shown.
  • N (responses per prompt in training data) = 4
    Number of diverse responses generated per prompt for the one-to-many data; influences the diversity ceiling.
  • beta (token penalty scale) = 1 for first token of each word, 0 otherwise
    Heuristic to avoid penalizing subword tokens; chosen without justification.
  • Reported temperatures = e.g., 0.5, 0.75, 1.0, 1.25 per method
    Only optimal temperatures per method are reported (Section 6), which is a test-set-dependent choice.
assumptions (5)
  • domain assumption GPT-4o with List Prompting produces valid, semantically distinct responses for a given prompt.
    Used to create all one-to-many training data (Section 5.1). If the teacher responses are not genuinely diverse, the method cannot teach diversity.
  • domain assumption SBERT pairwise similarity, averaged and subtracted from 1, is a valid measure of semantic diversity.
    Used as the main diversity metric (Section 5.2.1). No correlation with human judgments is shown.
  • domain assumption The percentage of responses rated below 6 by both GPT-4o and LLAMA 3 is a valid coherence metric.
    The coherence indicator (Section 5.2.2) is based on LLM ratings with a threshold, without human validation.
  • domain assumption Shannon entropy of GPT-4o-extracted persona attributes measures demographic bias.
    The persona-generation test (A.4) equates higher entropy with reduced bias; this ignores accuracy and realism of the attributes.
  • domain assumption Unlikelihood training on first tokens prevents out-of-vocabulary degeneration.
    The beta weighting in Section 4.3 assumes that penalizing only first tokens is sufficient to avoid OOV while suppressing wrong-k responses.
invented entities (1)
  • Possibility number k
    purpose: A conditioning token appended to the prompt that tells the model which of several possible responses to generate.
    The paper introduces this new input feature and claims it provides control (Section 4.2), but no experiment demonstrates that a fixed k reliably produces a consistent response type across prompts. It is a model input, not an observed phenomenon.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Linguistic Diversity of Large Language Models with Possibility Exploration Fine-Tuning." pith.science (2026). https://pith.science/paper/EL73KOFO

@misc{pith2026241203343,
  author       = {Pith},
  title        = {Pith review of: Improving Linguistic Diversity of Large Language Models with Possibility Exploration Fine-Tuning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EL73KOFO}},
  note         = {Machine review of arXiv:2412.03343}
}
read the original abstract

While Large Language Models (LLMs) have made significant strides in replicating human-like abilities, there are concerns about a reduction in the linguistic diversity of their outputs. This results in the homogenization of viewpoints and perspectives, as well as the underrepresentation of specific demographic groups. Although several fine-tuning and prompting techniques have been suggested to tackle the issue, they are often tailored to specific tasks or come with a substantial increase in computational cost and latency. This makes them challenging to apply to applications that demand very low latency, such as chatbots and virtual assistants. We propose Possibility Exploration Fine-Tuning (PEFT), a task-agnostic framework that enhances the text diversity of LLMs without increasing latency or computational cost. Given the same prompt, models fine-tuned with PEFT can simultaneously generate multiple diverse responses, each corresponding with a controllable possibility number. Experiments on dialogue and story generation tasks demonstrate that PEFT significantly enhances the diversity of LLM outputs, as evidenced by lower similarity between candidate responses. Since PEFT emphasizes semantic diversity over lexical diversity, it can also notably reduce demographic bias in dialogue systems. The implementations and datasets are available in our repository: https://github.com/mailong25/peft_diversity

Figures

Figures reproduced from arXiv: 2412.03343 by the authors.

Figure 1
Figure 1. An example of List Prompting for open￾domain dialogue generation generation by Hayati et al. (2023), we introduce List Prompting as a general framework for multi￾response generation using the following template: I want to <task description>. List a diverse set of <N> possible responses: An example of List Prompting for dialogue gen￾eration is shown in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. An simplified example of a PE training batch [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Persona demographic distributions extracted from 300 conversations with Mistral base and its fine-tuned [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Beyond Quality: Unlocking Diversity in Ad Headline Generation with Large Language Models

    cs.CL 2025-08 conditional novelty 5.0 of 10

    DIVER trains an LLM with synthetic data plus multi-objective RL to emit several diverse, style-tagged ad headlines in one pass, reporting online ADVV and CTR gains.

Reference graph

Works this paper leans on

40 extracted references · 7 canonical work pages · cited by 1 Pith paper

  1. [1]

    Siqi Bao, Huang He, Fan Wang, Hua Wu, Haifeng Wang, Wenquan Wu, Zhen Guo, Zhibin Liu, and Xinchao Xu. 2020. Plato-2: Towards building an open-domain chatbot via curriculum learning. arXiv preprint arXiv:2006.16779

    work page Pith review arXiv 2020

  2. [2]

    Siqi Bao, Huang He, Fan Wang, Hua Wu, Haifeng Wang, Wenquan Wu, Zhen Guo, Zhibin Liu, and Xinchao Xu. 2021. https://doi.org/10.18653/v1/2021.findings-acl.222 PLATO-2 : Towards building an open-domain chatbot via curriculum learning . In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 2513--2525, Online. Association for Co...

    doi:10.18653/v1/2021.findings-acl.222 2021

  3. [3]

    John Joon Young Chung, Ece Kamar, and Saleema Amershi. 2023. Increasing diversity while maintaining accuracy: Text data generation with large language models and human interventions. arXiv preprint arXiv:2306.04140

    arXiv 2023

  4. [4]

    Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. 2024. Qlora: Efficient finetuning of quantized llms. Advances in Neural Information Processing Systems, 36

    2024

  5. [5]

    Emily Dinan, Stephen Roller, Kurt Shuster, Angela Fan, Michael Auli, and Jason Weston. 2018. Wizard of wikipedia: Knowledge-powered conversational agents. arXiv preprint arXiv:1811.01241

    arXiv 2018

  6. [6]

    Wanyu Du, Jianqiao Zhao, Liwei Wang, and Yangfeng Ji. 2022. Diverse text generation via variational encoder-decoder models with gaussian process priors. arXiv preprint arXiv:2204.01227

    arXiv 2022

  7. [7]

    Esin Durmus, Karina Nyugen, Thomas I Liao, Nicholas Schiefer, Amanda Askell, Anton Bakhtin, Carol Chen, Zac Hatfield-Dodds, Danny Hernandez, Nicholas Joseph, et al. 2023. Towards measuring the representation of subjective global opinions in language models. arXiv preprint arXiv:2306.16388

    arXiv 2023

  8. [8]

    Angela Fan, Mike Lewis, and Yann Dauphin. 2018. Hierarchical neural story generation. arXiv preprint arXiv:1805.04833

    arXiv 2018

Show all 40 references
  1. [9]

    Isabel O Gallegos, Ryan A Rossi, Joe Barrow, Md Mehrab Tanjim, Sungchul Kim, Franck Dernoncourt, Tong Yu, Ruiyi Zhang, and Nesreen K Ahmed. 2023. Bias and fairness in large language models: A survey. arXiv preprint arXiv:2309.00770

    2023 arXiv

  2. [10]

    Aparna Garimella, Rada Mihalcea, and Akhash Amarnath. 2022. Demographic-aware language model fine-tuning as a bias mitigation technique. In Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International...

    2022

  3. [11]

    Karthik Gopalakrishnan, Behnam Hedayatnia, Qinlang Chen, Anna Gottardi, Sanjeev Kwatra, Anu Venkatesh, Raefer Gabriel, and Dilek Hakkani-Tur. 2023. Topical-chat: Towards knowledge-grounded open-domain conversations. arXiv preprint arXiv:2308.11995

    2023 arXiv

  4. [12]

    Shirley Anugrah Hayati, Minhwa Lee, Dheeraj Rajagopal, and Dongyeop Kang. 2023. How far can we extract diverse perspectives from large language models? criteria-based diversity prompting! arXiv preprint arXiv:2311.09799

    2023 arXiv

  5. [13]

    Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. 2019. The curious case of neural text degeneration. arXiv preprint arXiv:1904.09751

    2019 arXiv

  6. [14]

    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

    2023 arXiv

  7. [15]

    Preethi Lahoti, Nicholas Blumm, Xiao Ma, Raghavendra Kotikalapudi, Sahitya Potluri, Qijun Tan, Hansa Srinivasan, Ben Packer, Ahmad Beirami, Alex Beutel, et al. 2023. Improving diversity of demographic representation in large language models via collective-critiques and self-vo...

    2023 arXiv

  8. [16]

    Jing Yang Lee, Kong Aik Lee, and Woon-Seng Gan. 2023. An empirical bayes framework for open-domain dialogue generation. arXiv preprint arXiv:2311.10945

    2023 arXiv

  9. [17]

    Chia-Wei Liu, Ryan Lowe, Iulian V Serban, Michael Noseworthy, Laurent Charlin, and Joelle Pineau. 2016. How not to evaluate your dialogue system: An empirical study of unsupervised evaluation metrics for dialogue response generation. arXiv preprint arXiv:1603.08023

    2016 arXiv

  10. [18]

    Varvara Logacheva, Mikhail Burtsev, Valentin Malykh, Vadim Polulyakh, and Aleksandr Seliverstov. 2018. Convai dataset of topic-oriented human-to-chatbot dialogues. In The NIPS'17 Competition: Building Intelligent Systems, pages 47--57. Springer

    2018

  11. [19]

    Justus Mattern, Zhijing Jin, Mrinmaya Sachan, Rada Mihalcea, and Bernhard Sch \"o lkopf. 2022. Understanding stereotypes in language models: Towards robust measurement and zero-shot debiasing. arXiv preprint arXiv:2212.10678

    2022 arXiv

  12. [20]

    Nasrin Mostafazadeh, Nathanael Chambers, Xiaodong He, Devi Parikh, Dhruv Batra, Lucy Vanderwende, Pushmeet Kohli, and James Allen. 2016. A corpus and evaluation framework for deeper understanding of commonsense stories. arXiv preprint arXiv:1604.01696

    2016 arXiv

  13. [21]

    Nasrin Mostafazadeh, Michael Roth, Annie Louis, Nathanael Chambers, and James F Allen. 2017. Lsdsem 2017 shared task: The story cloze test. In 2nd Workshop on Linking Models of Lexical, Sentential and Discourse-level Semantics, pages 46--51. Association for Computational Linguistics

    2017

  14. [22]

    Jananee Muralidharan and Tiju Thomas. 2024. Deliberate problem-solving with a large language model as a brainstorm aid using a checklist for prompt generation

    2024

  15. [23]

    OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, Red Avila, Igor Babuschkin, Suchir Balaji, Valerie Balcom, Paul Baltescu, Haiming Bao, Mohammad Bavarian, Jeff ...

    2024 arXiv

  16. [24]

    Hannah Rashkin, Eric Michael Smith, Margaret Li, and Y-Lan Boureau. 2018. Towards empathetic open-domain conversation models: A new benchmark and dataset. arXiv preprint arXiv:1811.00207

    2018 arXiv

  17. [25]

    Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. arXiv preprint arXiv:1908.10084

    2019 arXiv

  18. [26]

    Shibani Santurkar, Esin Durmus, Faisal Ladhak, Cinoo Lee, Percy Liang, and Tatsunori Hashimoto. 2023. Whose opinions do language models reflect? In International Conference on Machine Learning, pages 29971--30004. PMLR

    2023

  19. [27]

    Claude Elwood Shannon. 1948. A mathematical theory of communication. The Bell system technical journal, 27(3):379--423

    1948

  20. [28]

    Zhihong Shao, Minlie Huang, Jiangtao Wen, Wenfei Xu, and Xiaoyan Zhu. 2019. Long and diverse text generation with planning-based hierarchical variational model. arXiv preprint arXiv:1908.06605

    2019 arXiv

  21. [29]

    Eric Michael Smith, Mary Williamson, Kurt Shuster, Jason Weston, and Y-Lan Boureau. 2020. Can you put it all together: Evaluating conversational agents' ability to blend skills. arXiv preprint arXiv:2004.08449

    2020 arXiv

  22. [30]

    Irene Solaiman and Christy Dennison. 2021. Process for adapting language models to society (palms) with values-targeted datasets. Advances in Neural Information Processing Systems, 34:5861--5873

    2021

  23. [31]

    Yixuan Su and Nigel Collier. 2022. Contrastive search is what you need for neural text generation. arXiv preprint arXiv:2210.14140

    2022 arXiv

  24. [32]

    Yixuan Su, Tian Lan, Yan Wang, Dani Yogatama, Lingpeng Kong, and Nigel Collier. 2022. A contrastive framework for neural text generation. Advances in Neural Information Processing Systems, 35:21548--21561

    2022

  25. [33]

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. 2023. https://arxiv.org/abs/2302.13971 Llama:...

    2023 arXiv

  26. [34]

    Ashwin K Vijayakumar, Michael Cogswell, Ramprasath R Selvaraju, Qing Sun, Stefan Lee, David Crandall, and Dhruv Batra. 2016. Diverse beam search: Decoding diverse solutions from neural sequence models. arXiv preprint arXiv:1610.02424

    2016 arXiv

  27. [35]

    Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2022. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171

    2022 arXiv

  28. [36]

    Sean Welleck, Ilia Kulikov, Stephen Roller, Emily Dinan, Kyunghyun Cho, and Jason Weston. 2019. Neural text generation with unlikelihood training. arXiv preprint arXiv:1908.04319

    2019 arXiv

  29. [37]

    Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, et al. 2024. Judging llm-as-a-judge with mt-bench and chatbot arena. Advances in Neural Information Processing Systems, 36

    2024

  30. [38]

    Ran Zmigrod, Sabrina J Mielke, Hanna Wallach, and Ryan Cotterell. 2019. Counterfactual data augmentation for mitigating gender stereotypes in languages with rich morphology. arXiv preprint arXiv:1906.04571

    2019 arXiv

  31. [39]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  32. [40]

    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 gl...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.