Pith. sign in

REVIEW 4 major objections 7 minor 14 references

Tuning-Free Personalized Alignment via Trial-Error-Explain In-Context Learning

T0 review · 4 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read TICL personalizes LLM writing with no fine-tuning, needing under 10 examples per user

desk verdict Useful method, honest limitations section, but the abstract’s headline number sits on the one result the authors themselves suspect is leaked. read the letter →

arxiv 2502.08972 v3 pith:DXQQPF6U submitted 2025-02-13 cs.CL cs.AI

classification cs.CLcs.AI
keywords TICLpersonalizedtextgenerationin-contextlearningtrial-error-explainLLM-as-a-judgestylisticalignmenttuning-freepersonalizationnegativesamplesandexplanations
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 proposes TICL, a tuning-free method that personalizes a language model's writing style without any parameter updates or large amounts of user data: it needs fewer than ten examples per user and only extra inference compute while building the prompt. TICL starts with a few-shot in-context-learning prompt, generates a candidate output, asks the model to explain how that output diverges stylistically from the user's text, and appends the candidate as a negative sample along with its explanation. This trial-error-explain loop repeats, and the best-performing augmented prompt is checkpointed and used at test time with no extra generation steps. On email, essay, and news-article writing tasks, TICL reports pairwise win rates up to 91.5% against the previous state of the art and outperforms other tuning-free baselines, with ablations showing that explanations carry most of the gain. If the evaluation is trustworthy, this means inference-only personalization can match or beat fine-tuned personalization, which matters for API-only models, privacy-sensitive users, and settings where per-user data is scarce.

What carries the argument

The central mechanism is the iteratively augmented in-context prompt, built from tuples of task, user text, and a set of model-generated negative outputs with explanations. In each trial-error-explain cycle, TICL samples one hold-out example, generates a candidate output from the current prompt, runs an explanation prompt that critiques the stylistic gap between the candidate and the user's reference text, and—if a validator flags the output as inconsistent—appends the candidate and its explanation to the prompt as a labeled negative example. Checkpointing selects the augmented dataset with the best validation win rate, so the final prompt carries both positive demonstrations and explicit "what not to do" guidance. This carries the argument because it turns preference optimization into context: no parameters change, and no extra decoding cost is paid at test time.

What would settle it

Ask five experienced editors or authorship-verification specialists to pick the more stylistically similar candidate on a random sample of the same pairwise comparisons that GPT-4o judged, then measure agreement; if agreement is near chance on the top-10 authors, the ~97% evaluator benchmark does not transfer and the central claim fails. A cheaper control is to run the same pairs with Claude as judge on GPT-4o outputs and vice versa and see whether the method's win-rate advantage reverses or disappears.

Watch

Extended reading notes

Core claim

The central claim is that the three stages of trial-and-error fine-tuning—behavior cloning, exploration, and preference learning—can be replaced by tuning-free prompt operations without sacrificing personalization quality. TICL substitutes few-shot in-context learning for supervised fine-tuning, generates one negative output at a time from the current prompt instead of a batch, and replaces preference optimization with prompt augmentation that appends the negative output and a model-generated explanation of its stylistic shortcomings. The paper reports that this method outperforms zero-shot, few-shot, chain-of-thought, and OPRO baselines, and beats DITTO, the previous fine-tuning-based state of the art, in 53.0% and 82.5% of pairwise LLM-as-judge comparisons with GPT-4o on CMCC and CCAT, rising to 79.5% and 91.5% with Claude 3 Sonnet. Against the author's actual text, Claude 3 Sonnet with TICL crosses the 50% threshold on both datasets, which the paper treats as near-indistinguishability under the judge. Ablations show that every component contributes, that removing explanations causes the largest drop, and that a model learns best from its own failure modes; lexical analysis indicates TICL reduces the model's bias toward formal connectives such as "additionally" and "therefore" in favor of the colloquial phrases found in user texts. The paper also notes a suspected data leakage for CCAT into Claude and therefore focuses subsequent analysis on CMCC.

Load-bearing premise

The central result rests on the assumption that GPT-4o, used as the judge of stylistic similarity, is both accurate and unbiased; if the judge secretly favors its own generation style or has memorized the benchmark authors, the reported win rates—especially the 91.5% against DITTO—would not reflect true user alignment.

Editorial extensions

If this is right

  • Personalized text generation can be delivered without per-user fine-tuning, making it feasible for API-only models and for users who cannot or will not share large amounts of data.
  • Negative samples with explanations are the main driver of style adaptation; the reported ablation shows explanations alone can account for up to 77% of the gain over plain few-shot prompting.
  • TICL prompts are not portable across models: swapping the model that generated the negative samples and explanations degrades performance, so personalization must be recomputed for each target model.
  • The method inherits a hard dependency on long-context understanding; the paper reports that smaller models imitate the negative examples rather than avoid them, so TICL-style personalization will become more broadly applicable as long-context models improve.
  • Front-loading the trial-error loop into the prompt means no extra decoding steps at test time, so the extra cost is a one-time per-user prompt construction cost rather than a per-generation cost.

Reading between the lines

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

  • Editorial inference: the headline 91.5% number is the least stable result in the paper, because the authors themselves suspect Claude has seen CCAT articles and DITTO runs on a much smaller model; a fair stress test would re-run on freshly collected personal writing with a judge from a different model family.
  • Editorial inference: the trial-error-explain loop is a generic recipe: any task where a model can compare a candidate against a reference and critique the gap could be personalized this way, so code style, translation style, or formatting preferences are immediately testable extensions.
  • Editorial inference: because prompts are model-specific and long-context-dependent, the practical next step is compressing the augmented prompt into a style guide and checking whether the win-rate gains survive; the paper's limitations section indicates this direction but does not test it.
  • Editorial inference: a cross-model judge swap—Claude judging GPT-4o outputs and vice versa—would estimate how much of the win rate is true stylistic alignment versus judge self-preference, and would be a cheap, decisive control experiment.
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

4 major / 7 minor

Summary. The paper proposes TICL, a tuning-free method that personalizes large language models for text generation by iteratively expanding an in-context learning prompt with model-generated negative samples and explanations, drawing an explicit analogy to the fine-tuning method TEFT/DITTO. The method is evaluated on two authorship datasets (CMCC and CCAT) with GPT-4o and Claude 3 Sonnet, using a GPT-4o pairwise LLM-as-a-judge to compare stylistic similarity against DITTO outputs and against the author's own text. The authors report win rates up to 91.5% against DITTO, run ablations showing that explanations are the most important component, and provide lexical analyses linking the performance differences to the models' handling of formal versus colloquial phrases.

Significance. If the reported win rates hold, TICL would be a practically valuable contribution: it achieves personalization with fewer than ten examples per user, requires no parameter updates, and works with API-only models, which addresses a real limitation of fine-tuning-based personalization. The method is clearly described, the ablation design credibly isolates the contribution of each component, the prompt templates and processed data are shared, and the authors are candid about the method's dependence on long-context understanding and about the per-user prompt storage cost. However, the central quantitative claim is currently tied to an evaluation setup that the authors themselves partially call into question, so the significance is conditional on a stronger or more carefully caveated evaluation.

major comments (4)
  1. [Abstract and §4.1 (Table 2)] The abstract's headline 'up to 91.5% against the previous state-of-the-art' is the CCAT win rate of Claude 3 Sonnet TICL versus DITTO in Table 2. In §4.1 the authors state: 'Given Claude's strong performance even in the zero-shot setting against DITTO, we suspect there may be some leakage of CCAT data into Claude. Therefore, we focus our analysis henceforth mainly on CMCC.' Despite this explicit suspicion, the abstract and Table 2 retain the 91.5% value without caveat, and no contamination-controlled experiment is presented for CCAT. This is load-bearing because if Claude has memorized CCAT articles, the pairwise judge's preference may reflect memory of the source text rather than the TICL procedure. Please either remove the CCAT result from the headline, report it with an explicit caveat, or add a contamination test (for example, n-gram overlap with pretraining data, generation on held-out CCAT-style prompts, or evaluation on a dataset with post-training cutoff dates).
  2. [§3.2 and Appendix A.1] The evaluation selects the top 10 authors per dataset based on the judge's own accuracy (Table 1 and Table 6), and each author contributes only three test prompts (Appendix A.1). This makes the reported averages conditional on authors for whom the judge is already known to work, and the per-author test set is very small. The paper's claim of personalized alignment 'with fewer than 10 examples per user' would be more credible if results were reported for all authors or a random subset, with per-prompt variance; as written, the selection procedure limits the generality of the headline win rates and makes the method's average performance hard to interpret for a typical new user.
  3. [§3.2, Table 1] The LLM-as-a-judge benchmarking in Table 1 validates the judge on human-written author texts (author versus distractor), not on model-generated candidates of the kind TICL and the baselines produce. Since the judge and one of the two main generators are both GPT-4o, there is a risk that the judge prefers outputs that resemble GPT-4o's default style rather than the target author's style; the reported ~97% accuracy on human texts does not rule this out. A concrete test would be to evaluate a sample with a different judge model (for example, Claude 3 Sonnet or GPT-4o mini) or with an embedding-based stylistic similarity metric and to report agreement; if this is infeasible, the 'stylistic similarity' claim should be tempered.
  4. [Table 2, §3.3, Appendix A.2] The comparison against DITTO conflates method with base model and with output filtering. TICL runs on GPT-4o or Claude 3 Sonnet, while DITTO is trained on Mistral 7B Instruct, a much smaller model, and Appendix A.2 explains that DITTO outputs containing template artifacts or repetitive content were rejected until clean samples were obtained. The phrase 'previous state-of-the-art' should therefore be framed as 'TICL on a strong API model beats DITTO on Mistral 7B with output filtering,' or the authors should include a DITTO variant using a comparable base model, otherwise the comparison does not isolate the benefit of the TICL procedure.
minor comments (7)
  1. [Abstract] The abstract should specify that the 91.5% figure is for Claude 3 Sonnet on CCAT and that the authors themselves suspect pretraining leakage in that setting; as written, the 'up to' construction obscures the caveat.
  2. [§4.1] The sentence expressing the leakage suspicion should be moved to or duplicated in the Limitations section and should be reflected in the abstract and in the caption of Table 2.
  3. [Algorithm 1 and §2.3] Algorithm 1 returns DTICL after L iterations, but §2.3 describes checkpointing by validation performance; the algorithm should include the checkpoint and selection step for reproducibility.
  4. [Table 3] The row labeled '− Negative samples & Expl.' is described in the text as equivalent to the few-shot baseline; the caption should state this explicitly so the reader can see that the ablation's final row is a baseline.
  5. [§4.4] The text 'FRE(B\A=77.21' is missing a closing parenthesis, and 'more frequenty' is a typo for 'frequently'.
  6. [Appendix A.2] The phrase 'To give these approaches the best chances possible against our approaches' should be rephrased; the antecedent of 'our approaches' is unclear and it should be 'against our approach' or similar.
  7. [Table 8 footnote] The qualification that TICL does not significantly outperform CoT for Claude 3 Sonnet on CCAT appears only in the table footnote; this should be stated in the main text since it directly qualifies the abstract's 'outperforms' wording.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: TICL is an empirical prompting method benchmarked against external datasets and baselines; the CCAT leakage caveat is an evaluation-validity concern, not a self-referential reduction.

full rationale

This paper does not derive a formal mathematical result; it proposes an inference-time procedure and measures win rates on two external authorship datasets (CMCC and CCAT) against DITTO and other baselines. The TICL prompt is expanded by model-generated negative samples and explanations, but there is no fitted parameter later relabeled as a prediction, and no equation in which the reported output reduces to the input by construction. The evaluation uses GPT-4o as a judge, which is the same model family as the generator in some conditions, but the judge was separately benchmarked to roughly 97% authorship-attribution accuracy on the top-10 authors, so the win rates are not defined as the judge's own preferences by construction. Several citations are to prior work by overlapping authors (e.g., Liu et al. 2024; Liu and May 2024) in support of using model-based style judgment, but the same claim is also supported by external references and is not the load-bearing derivation of TICL. The clearest weakness is not circularity: Section 4.1 states, "Given Claude's strong performance even in the zero-shot setting against DITTO, we suspect there may be some leakage of CCAT data into Claude. Therefore, we focus our analysis henceforth mainly on CMCC," yet the abstract reports the 91.5% CCAT result without that caveat. That is a benchmark-contamination or soundness concern, not a self-referential reduction; the CMCC results and ablations remain independent evidence. Overall, no circular step satisfying the definitional, fitted-input, or self-citation-chain tests can be exhibited from the paper's own equations or construction.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

TICL introduces no new physical or ontological entities; it is a prompt augmentation procedure. The validator flag and negative samples are model outputs, not independent constructs. The free parameters are hand-chosen hyperparameters of the method and evaluation, not numbers fit to data. The axioms are domain assumptions about the validity of LLM-style judgment and the efficacy of self-generated negatives.

free parameters (4)
  • Number of TICL iterations (epochs) = 4
    Set to match DITTO's four epochs; chosen by hand, not optimized, and the paper does not report a sensitivity analysis.
  • Number of initial in-context examples per author = 6-7
    Determined by the 7/2/3 train/val/test split of the 12 available samples per author; this is a data constraint rather than a tuned hyperparameter.
  • Number of author examples used by the LLM judge = 5
    Chosen to increase judge accuracy from 76.1% (1 example) to 89.5% (5 examples) on CMCC, based on Table 7; this is a hand-selected evaluation setting, not a method parameter.
  • Number of comparison pairs sampled per author = 40 (vs DITTO), 30 (vs author)
    Subsampled from the generated outputs to control cost; the random seed is not reported, which adds variance to the win-rate estimates.
assumptions (5)
  • domain assumption LLM-as-a-judge with GPT-4o can reliably judge stylistic similarity between texts.
    The paper validates this on authorship attribution, reaching about 97% accuracy for top-10 authors, but the judge is not independently verified for the pairwise comparisons of generated texts, and the same model family is used to generate the outputs.
  • domain assumption Stylistic consistency can be decomposed into the eight listed dimensions (length, format, paragraph structure, sentence structure, punctuation, syntax, voice, diction).
    This taxonomy is used in the generation prompts, the explanation prompt, and the judge prompt; it is a modeling choice and is not demonstrated to be exhaustive or unique.
  • domain assumption Negative samples generated by the target model itself provide a useful error signal for in-context learning.
    This is the core TICL mechanism. Ablations show that removing negative samples and explanations hurts performance, but there is no independent evidence that the negative samples are not simply memorized or that they generalize beyond the specific test authors.
  • domain assumption The validation set can be used to select the best TICL prompt version (checkpointing) without overfitting the test set.
    The paper uses win rates on a small validation set (two examples per author) to pick the best prompt iteration; this is a standard practice but the tiny validation size makes the selection unstable.
  • domain assumption DITTO's outputs, generated with Mistral 7B, fairly represent the previous state of the art for personalized text generation.
    The comparison against DITTO mixes model scale and method, because DITTO uses a much smaller base model. A fine-tuned GPT-4o or Claude 3 Sonnet baseline is absent, so the reported advantage may be part base-model capability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tuning-Free Personalized Alignment via Trial-Error-Explain In-Context Learning." pith.science (2026). https://pith.science/paper/DXQQPF6U

@misc{pith2026250208972,
  author       = {Pith},
  title        = {Pith review of: Tuning-Free Personalized Alignment via Trial-Error-Explain In-Context Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DXQQPF6U}},
  note         = {Machine review of arXiv:2502.08972}
}
read the original abstract

Language models are aligned to the collective voice of many, resulting in generic outputs that do not align with specific users' styles. In this work, we present Trial-Error-Explain In-Context Learning (TICL), a tuning-free method that personalizes language models for text generation tasks with fewer than 10 examples per user. TICL iteratively expands an in-context learning prompt via a trial-error-explain process, adding model-generated negative samples and explanations that provide fine-grained guidance towards a specific user's style. TICL achieves favorable win rates on pairwise comparisons with LLM-as-a-judge up to 91.5% against the previous state-of-the-art and outperforms competitive tuning-free baselines for personalized alignment tasks of writing emails, essays and news articles. Both lexical and qualitative analyses show that the negative samples and explanations enable language models to learn stylistic context more effectively and overcome the bias towards structural and formal phrases observed in their zero-shot outputs. By front-loading inference compute to create a user-specific in-context learning prompt that does not require extra generation steps at test time, TICL presents a novel yet simple approach for personalized alignment.

Figures

Figures reproduced from arXiv: 2502.08972 by the authors.

Figure 1
Figure 1. Top: The author’s text contains rhetorical questions and more colloquial and informal phrases. Middle: Few-shot in-context learning is insufficient for adapting to an author’s style and model biases for formal phrases persist. Bottom: TICL helps overcome model biases and applies the author’s style more consistently. Model outputs are with gpt-4o-2024-0806 prior work on personalized text generation either rely on lar… view at source ↗
Figure 2
Figure 2. TICL methodology overview. Instead of SFT, TICL starts with a few-shot ICL for behavior cloning. Then, TICL repeatedly generates an output for a task and a corresponding explanation that critiques the stylistic difference between the output and the user’s text. If the output is considered not stylistically consistent, it and its explanation are added to the prompt. (NLL) loss to learn from DT to fine-tune a pre￾trai… view at source ↗
Figure 3
Figure 3. Simplified illustration of our evaluation setup [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Prompt template E for generating explanations on the difference between the reference output and the generated output. You are a stylistically consistent writer. Below are examples that exemplify your writing style. # Writing Task Example 1 { example["task"] } ## Your …
Figure 5
Figure 5. Figure 5: Prompt template P for the few-shot in-context learning baseline [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: Prompt template P for TICL. It builds on [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 7
Figure 7. Figure 7: Prompt template for evaluating which output is more stylistically similar to the target author’s writing [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Prompt template for generating style guides that will be used for Chain-of-Thought guided writing in [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]
Figure 9
Figure 9. Figure 9: Prompt template for Chain-of-Thought-guided writing. The [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]
Figure 10
Figure 10. Figure 10: OPRO optimization prompt. The instruction with the highest score is used for the OPRO writing prompt [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: OPRO writing prompt. The blue text is the instruction that is selected from the prompt optimization process from [PITH_FULL_IMAGE:figures/full_fig_p019_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 5 canonical work pages

  1. [2]

    In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 13376–13390, Miami, Florida, USA

    TinyStyler: Efficient few-shot text style trans- fer with authorship embeddings. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 13376–13390, Miami, Florida, USA. Association for Computational Linguistics. Joel Jang, Seungone Kim, Bill Yuchen Lin, Yizhong Wang, Jack Hessel, Luke Zettlemoyer, Hannaneh Hajishirzi, Yejin Choi,...

  2. [4]

    In The Twelfth Inter- national Conference on Learning Representations

    Prometheus: Inducing fine-grained evaluation capability in language models. In The Twelfth Inter- national Conference on Learning Representations . Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yu- taka Matsuo, and Yusuke Iwasawa. 2022. Large lan- guage models are zero-shot reasoners. Advances in neural information processing systems , 35:22199– 22213. ...

  3. [5]

    Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E

    Longlamp: A benchmark for personalized long-form text generation. Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. 2023. Effi- cient memory management for large language model serving with pagedattention. In Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Princ...

  4. [6]

    arXiv preprint arXiv:2309.00267

    Rlaif: Scaling reinforcement learning from human feedback with ai feedback. arXiv preprint arXiv:2309.00267. David D Lewis, Yiming Yang, Tony Russell-Rose, and Fan Li. 2004. Rcv1: A new benchmark collection for text categorization research. Journal of machine learning research, 5(Apr):361–397. Cheng Li, Mingyang Zhang, Qiaozhu Mei, Weize Kong, and Michael...

  5. [9]

    Advances in Neu- ral Information Processing Systems, 36

    Direct preference optimization: Your language model is secretly a reward model. Advances in Neu- ral Information Processing Systems, 36. 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...

  6. [11]

    In Pro- ceedings of the 7th Workshop on Representation Learning for NLP, pages 249–268, Dublin, Ireland

    Same author or just same topic? towards content-independent style representations. In Pro- ceedings of the 7th Workshop on Representation Learning for NLP, pages 249–268, Dublin, Ireland. Association for Computational Linguistics. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought ...

  7. [12]

    differentiation

    Large language models as optimizers. In The Twelfth International Conference on Learning Representations. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2023. React: Synergizing reasoning and acting in language models. In International Conference on Learning Representations (ICLR). Qinyuan Ye, Mohamed Ahmed, Re...

  8. [13]

    arXiv preprint arXiv:2306.05685

    Judging llm-as-a-judge with mt-bench and chatbot arena. arXiv preprint arXiv:2306.05685. Pei Zhou, Hyundong Cho, Pegah Jandaghi, Dong-Ho Lee, Bill Yuchen Lin, Jay Pujara, and Xiang Ren

Show all 14 references
  1. [14]

    Write a news article that starts with the following sentence : article’s first sentence

    Reflect, not reflex: Inference-based common ground improves dialogue response quality. In Pro- ceedings of the 2022 Conference on Empirical Meth- ods in Natural Language Processing, pages 10450– 10468, Abu Dhabi, United Arab Emirates. Associa- tion for Computational Linguistic...

  2. [908]

    Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, and Denny Zhou

    IEEE. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, and Denny Zhou. 2022. Rationale- augmented ensembles in language models. arXiv preprint arXiv:2207.00747. Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A. Smith, Daniel Khashabi, and Hannaneh Hajish...

  3. [2008]

    Political Analysis, 16(4):372–403

    Fightin’ words: Lexical feature selection and evaluation for identifying the content of political con- flict. Political Analysis, 16(4):372–403. Sheshera Mysore, Zhuoran Lu, Mengting Wan, Longqi Yang, Steve Menezes, Tina Baghaee, Em- manuel Barajas Gonzalez, Jennifer Neville, ...

  4. [2022]

    arXiv preprint arXiv:2204.09391

    You are what you write: Preserving privacy in the era of large language models. arXiv preprint arXiv:2204.09391. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christo- pher D Manning, Stefano Ermon, and Chelsea Finn

  5. [2023]

    arXiv preprint arXiv:2310.11564

    Personalized soups: Personalized large lan- guage model alignment via post-hoc parameter merg- ing. arXiv preprint arXiv:2310.11564. Albert Q Jiang, Alexandre Sablayrolles, Arthur Men- sch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Len...

  6. [2024]

    Ritam Dutt, Kasturi Bhattacharjee, Rashmi Gangadhara- iah, Dan Roth, and Carolyn Rose

    Speechworthy instruction-tuned language models. Ritam Dutt, Kasturi Bhattacharjee, Rashmi Gangadhara- iah, Dan Roth, and Carolyn Rose. 2022. PerKGQA: Question answering over personalized knowledge graphs. In Findings of the Association for Compu- tational Linguistics: NAACL 20...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.