Pith. sign in

REVIEW 4 major objections 5 minor 37 references

From We to Me: Theory Informed Narrative Shift with Abductive Reasoning

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Pith's one-line read Abduction-guided rewriting makes LLMs shift a story from collectivistic to individualistic framing while preserving fidelity.

desk verdict A genuinely new task and a clever abduction-plus-LLM pipeline, but the headline numbers rest on an unvalidated LLM self-assessment loop; human evaluation is needed before the claims can be trusted. read the letter →

arxiv 2603.03320 v2 pith:R5PAGU4H submitted 2026-02-10 cs.CL cs.AI

classification cs.CLcs.AI
keywords narrativeshiftindividualism-collectivismabductivereasoningneuroSymbolicNLPLLMtexttransformationculturalframingstyletransfersemanticfidelity
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 tries to show that large language models fail at transforming the cultural orientation of a story unless they are told exactly which story elements to change. It proposes a neurosymbolic pipeline that learns rules from a corpus of target narratives, uses abductive reasoning to pick the specific chunks that carry collectivistic or individualistic framing, and only then asks an LLM to rewrite those chunks. On a corpus of 118 narratives, this approach outperformed zero-shot prompting for both shift directions across several LLMs. The headline numbers: for collectivistic-to-individualistic shift with GPT-4o, 55.88% better narrative shift and 40.4% better semantic fidelity (KL divergence) than the zero-shot baseline.

What carries the argument

The load-bearing object is an abductive selector over an annotated logic program. Rules learned from target-orientation stories link chunk features (e.g., 'collective action', 'uniqueness') to narrative orientation; the abduction step then proposes the minimal set of feature reassignments that raises the story's similarity score to the target corpus, and only those chunks are sent to the LLM for rewriting. A 40-question LLM-answered diagnostic survey (20 individualistic, 20 collectivistic features) supplies the feature annotations and the scoring metric.

What would settle it

Have a panel of human raters answer the same 40 diagnostic questions about the original and transformed stories. If the human scores do not move in the same direction as the LLM scores—or if they fail to separate the paper's own corpus into its 90 individualistic and 28 collectivistic stories—the central claim lacks independent support. A simpler check: run the diagnostic on stories with known cultural origins and see whether it classifies them better than chance.

Watch

Extended reading notes

Core claim

The paper claims narrative shift is doable when an LLM is told exactly which story elements to change. The authors formalize the task as abduction over confidence-annotated logic facts: each story chunk is described by narrative features, and the system selects the smallest set of feature changes that would make the story resemble a target corpus. An LLM then rewrites only those chunks. On 118 stories, the approach outperformed zero-shot prompting in most model-direction pairs; for GPT-4o on collectivistic-to-individualistic shift the gain was 55.88% in diagnosis score while keeping KL divergence to the original 40.4% lower (better) than the baseline.

Load-bearing premise

The load-bearing premise is that the 40-question LLM-based diagnostic survey validly measures how individualistic or collectivistic a narrative is; the same survey both decides which chunks to change and judges whether the change worked, and it is never checked against human ratings or an external benchmark.

Editorial extensions

If this is right

  • Zero-shot LLMs struggle with narrative shift; without guidance, they often leave collectivistic framing untouched or drift into incoherent rewrites.
  • The abduction-guided pipeline improves narrative shift over zero-shot prompting across multiple LLMs and in both directions, with one reported exception (DeepSeek-R1 on individualistic-to-collectivistic shift).
  • The method preserves more of the original story: KL divergence to the source is consistently lower than for the zero-shot baseline.
  • It is efficient: only a median of 32.11% of tokens are rewritten, and the number of LLM calls grows linearly with the size of the identified segments.
  • No task-specific fine-tuning is required; the same rule-learning plus fixed-prompt pattern works on both reasoning and non-reasoning LLMs.

Reading between the lines

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

  • The extract-then-rewrite pattern could transfer to other theory-grounded narrative dimensions (e.g., long-term vs. short-term time orientation, tolerance for uncertainty) wherever a diagnostic survey exists.
  • Replacing the LLM-based diagnostic with human raters or a separately trained classifier would test whether the method's gains are independent of the scoring instrument.
  • The median 32% token touch rate suggests the approach may double as a controlled-editing tool for style transfer generally, pinpointing exactly which sentences carry the attribute.
  • Because the same diagnostic both selects chunks and scores the outcome, a human-validation study would clarify how much of the reported improvement reflects genuine narrative shift versus a self-consistent scoring loop.
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 / 5 minor

Summary. The paper proposes a neurosymbolic approach to narrative shift along the individualism/collectivism dimension. The method uses a 40-item LLM-administered diagnostic survey to annotate story chunks with narrative features, learns weighted logical rules from a corpus of target-orientation stories, and then performs iterative abductive reasoning to select chunks for LLM-based transformation. The authors report that for GPT-4o the abduction-guided method outperforms a zero-shot baseline by 55.88% for collectivistic-to-individualistic shift while also showing a 40.4% improvement in KL divergence-based content similarity. Similar trends are reported for Grok-4, Llama-4, and DeepSeek-R1.

Significance. If the empirical claims were sound, the paper would offer a useful, fine-tuning-free method for controlling cultural framing in LLM-generated text, and the formal abduction framework is a plausible way to make targeted, explainable edits. However, the central results rest on an unvalidated self-referential diagnostic: the same 40-item LLM survey drives the chunk selection in the transformation pipeline and serves as the primary outcome metric. Without external validation (human ratings, established instruments, or a held-out test set), the reported improvements may be artifacts of the LLM optimizing its own survey responses. The fidelity metric is also too weak to support the 'semantic similarity' claim. The paper is clearly written and the pipeline is reproducible from the appendices, but the central evidence does not currently support the stated conclusions.

major comments (4)
  1. [Section 5 (Algorithm 1) and Section 6 (Evaluation Metrics)] The evaluation is circular. The 40-item diagnostic is used both to construct the observations for abduction (diagnosis(s) in Algorithm 1, line 3) and as the outcome metric ('diagnosis score'). Since the LLM prompt explicitly asks to make the story 'more [individualistic/collectivistic]' and the diagnostic items explicitly name the target features (Appendix B, e.g., 'Does the narrative present the self as stable and independent...?'), the method can move its own score by superficial rewrites. No human validation, external benchmark, or calibration study is provided. The paper's own observation that 9/28 collectivistic stories were already diagnosed as individualistic and 6 as neutral (Section 6) suggests poor discriminative validity of the diagnostic on the source corpus.
  2. [Section 5, Phase 1 and Section 6, Setup] No train/test split is described. The paper introduces a training corpus S_train for rule learning but never specifies how the 118 stories are divided between rule learning and evaluation, and no cross-validation is reported. Consequently, the rules could be overfit to the evaluation stories, and the reported improvements may not generalize. The manuscript must state an explicit split (or k-fold procedure) and report results on a held-out set.
  3. [Section 6, Eq. (6)] The KL divergence metric uses unigram token distributions with additive smoothing (alpha=1e-5). This measures surface lexical overlap, not semantic fidelity. The claim of '40.4% improvement in semantic similarity' is therefore not supported; a story can be lexically close yet change its meaning, and vice versa. Human evaluation of meaning preservation, or a validated semantic similarity metric, is required before conclusions about fidelity can be drawn.
  4. [Appendix A / Proposition 5.1] The proof of Proposition 5.1 is invalid. It defines N_c as the sum of token counts in chunks and compares it to k, the number of LLM transformation calls. The proof asserts that k + N_c cannot exceed N' (total tokens) because 'LLM calls in addition to non-identified tokens' cannot exceed the total tokens. This mixes incommensurable quantities (calls vs. tokens) and does not establish the claimed bound. The subsequent token-usage result (Section 6, R^2=0.843) does not repair the proof.
minor comments (5)
  1. [Abstract/Introduction] 'we learns rules' is a typo; should be 'we learn rules.'
  2. [Section 2] 'Annotations are are elements' should be 'Annotations are elements.'
  3. [Section 6, Narrative Diagnosis] The text says 'median of the 20 diagnosis scores' but the survey contains 40 questions (20 individualistic + 20 collectivistic). Clarify how the 20 scores are derived from the 40 items.
  4. [Figure 3] The stability plot is not described in the caption; please specify what is plotted, how the 10 runs differ, and which stories are shown.
  5. [General] No confidence intervals or significance tests are reported for the headline percentages (Tables 1-2); adding these would help assess the robustness of the comparisons.

Circularity Check

2 steps flagged · score 7.0 of 10

Closed evaluation loop: the same LLM diagnostic selects chunks, picks the best iteration, and defines narrative shift.

  1. self definitional [Definition 4.1; Algorithm 1 line 3; Section 6 'Narrative Diagnosis' and 'Evaluation Metrics']
    "Given a candidate story, we use an LLM to answer the survey to obtain ratings of the same story 10 times. ... these ratings are normalized to form the annotations in the output of diagnosis(s) introduced in Section 4. First, to capture the narrative of a story we define diagnosis score– given a narrative, we use a diagnosis to evaluate how individualistic or collectivistic the narrative is based on a survey."

    The same diagnosis(s) function is both the optimization signal and the outcome measure. Algorithm 1 starts each iteration with O_t ← diagnosis(s_t); the abduced explanation E maximizes σ, a corpus-similarity score computed from s_feat/c_feat atoms produced by diagnosis. The reported 'narrative shift' is the median of the same LLM-survey ratings. Thus the 55.88% improvement is a movement in the authors' unvalidated 40-item LLM survey, not an externally validated cultural shift. The method is optimizing the LLM to its own survey; no human rating or established instrument breaks the loop.

  2. fitted input called prediction [Algorithm 1, line 12; Section 6 'Evaluation Metrics']
    "12:t∗ ← arg max t{median(Ot)} ... we compute the median of the 20 diagnosis scores from the narrative diagnosis survey (explained in the Section 2) on transformed stories and report the average improvement over the original in the Table 1 and 2."

    The reported 'Abduction (Ours)' result is not a held-out prediction: the algorithm selects the iteration with the highest median diagnosis score, and that same median diagnosis score is the paper's success metric. This is optimizing on the evaluation metric. The zero-shot baseline is not given the arg-max selection step, so the headline improvement is inflated by construction rather than being an unbiased estimate of narrative-shift quality.

full rationale

The formal abduction machinery (Definitions 4.1-4.3, Algorithm 1) is internally coherent, and the comparison against a zero-shot baseline is real; self-citations to PyReason and prior abduction work are not load-bearing for the empirical claim. The circularity is in the evaluation loop: diagnosis(s), implemented as an LLM answering the authors' 40-item survey, is used (a) to build observations O_t that drive chunk selection, (b) to define the corpus-similarity objective σ via s_feat/c_feat annotations, (c) to choose the best iteration via arg max median(O_t), and (d) as the sole outcome metric ('diagnosis score'). Therefore the central claim—abduction-guided stories shifted narrative while preserving fidelity—rests on an internal score that the method is explicitly designed to increase. The paper's diagnostic note that 9/28 collectivistic stories scored individualistic and 6 neutral further indicates poor calibration, but even without it, the closed loop means the 55.88%/40.4% headline numbers cannot be read as independent confirmation. The KL-fidelity result is also weakened by the prompt's instruction to change only the selected segment, making low unigram KL partly a construction artifact; I treat that as a secondary validity concern rather than a separate circular step. Overall: partial circularity in the central claim, score 7.

Assumptions & free parameters 3 free parameters · 6 assumptions · 1 invented entities

The method relies on a fitted confidence table, a tuned hyperparameter, and a purpose-built unvalidated diagnostic instrument; the evaluation loop closes around this instrument.

free parameters (3)
  • conf(f, r) = not reported; fraction of training stories with feature f at rating r
    Learned from the training corpus in Phase 1 (§5) and used to form corpus-sim rules; this is the main fitted quantity of the method.
  • ψ (number of abduced features) = 2 for C→I, 3 for I→C
    Hyperparameter selected from sensitivity experiments on the same dataset (§6); no held-out validation.
  • target annotation τ = not specified
    Passed to llm_transform (§5); the paper does not state how τ is determined.
assumptions (6)
  • domain assumption The 40-item diagnostic survey measures narrative individualism/collectivism
    Used to generate diagnosis(s), which drives both abduction and evaluation; no human validation or external benchmark (§2, §6).
  • domain assumption LLM survey responses are stable and valid proxies for human narrative orientation
    Only 10-run stability is reported (Figure 3); validity is not established.
  • standard math Annotated logic programming semantics (Kifer & Subrahmanian) correctly implement the deduction
    The Γ* closure in Section 4 relies on this background; PyReason is cited but no version/commit.
  • domain assumption Chunks are coherent narrative units that can be modified independently without breaking the story
    Chunking is not described; the iterative transformation assumes isolated edits preserve coherence (§5).
  • ad hoc to paper The 118-story corpus can be used for both rule learning and evaluation without leakage
    No train/test split is described; the same corpus appears to be used for both phases (§5-6).
  • domain assumption KL divergence over smoothed unigram distributions measures semantic fidelity
    Used as fidelity metric (§6); unigram KL is a weak proxy for meaning preservation.
invented entities (1)
  • 40-item narrative diagnostic survey (20 individualistic + 20 collectivistic features)
    purpose: Defines the narrative-orientation score used both to guide abductive transformation and to evaluate success
    The survey is authored by the team, not validated against human raters or an existing cross-cultural instrument; it is the sole arbiter of narrative shift in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From We to Me: Theory Informed Narrative Shift with Abductive Reasoning." pith.science (2026). https://pith.science/paper/R5PAGU4H

@misc{pith2026260303320,
  author       = {Pith},
  title        = {Pith review of: From We to Me: Theory Informed Narrative Shift with Abductive Reasoning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R5PAGU4H}},
  note         = {Machine review of arXiv:2603.03320}
}
read the original abstract

Effective communication often relies on aligning a message with an audience's narrative and worldview. Narrative shift involves transforming text to reflect a different narrative framework while preserving its original core message--a task we demonstrate is significantly challenging for current Large Language Models (LLMs). To address this, we propose a neurosymbolic approach grounded in social science theory and abductive reasoning. Our method automatically extracts rules to abduce the specific story elements needed to guide an LLM through a consistent and targeted narrative transformation. Across multiple LLMs, abduction-guided transformed stories shifted the narrative while maintaining the fidelity with the original story. For example, with GPT-4o we outperform the zero-shot LLM baseline by 55.88% for collectivistic to individualistic narrative shift while maintaining superior semantic similarity with the original stories (40.4% improvement in KL divergence). For individualistic to collectivistic transformation, we achieve comparable improvements. We show similar performance across both directions for Llama-4, and Grok-4 and competitive performance for Deepseek-R1.

Figures

Figures reproduced from arXiv: 2603.03320 by the authors.

Figure 1
Figure 1. Abduction-guided LLM transformation correctly identi [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Abduction-guided Transformation (Our Approach). [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Stability of Social-Science LLM diagnostics over 10 runs [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (20 more)
Figure 5
Figure 5. Figure 5: The KL divergence across the stories for C [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 8
Figure 8. Figure 8: Total tokens of identified segments vs Number of LLM [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 6
Figure 6. Figure 6: The KL divergence across the stories for I [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Hyperparameter sensitivity. Left: C→I, Right: I→C. diagnosed as neutral narratives. All of these stories were successfully transformed into the target narrative. Addition￾ally, our abduction-guided approach demonstrates efficiency by targeting only essential narrative …
Figure 9
Figure 9. Figure 9: Complete individualistic diagnostic questions: 20 questions assessing individualistic narrative orientation. [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Complete collectivistic diagnostic questions: 20 questions assessing collectivistic narrative orientation. [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Transformation prompt template. Placeholders [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 12
Figure 12. Figure 12: The median of diagnosis ratings for GPT-4o across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]
Figure 13
Figure 13. Figure 13: The median of diagnosis ratings for Grok-4 across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]
Figure 14
Figure 14. Figure 14: The median of diagnosis ratings for Deepseek-R1 across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p013_14.png]
Figure 15
Figure 15. Figure 15: The median of diagnosis ratings for Llama across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p014_15.png]
Figure 16
Figure 16. Figure 16: The mode of diagnosis ratings for GPT-4o across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p014_16.png]
Figure 17
Figure 17. Figure 17: The mode of diagnosis ratings for Grok-4 across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p014_17.png]
Figure 18
Figure 18. Figure 18: The mode of diagnosis ratings for Deepseek-R1 across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p014_18.png]
Figure 19
Figure 19. Figure 19: The mode of diagnosis ratings for Llama across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p015_19.png]
Figure 20
Figure 20. Figure 20: The mean of diagnosis ratings for GPT-4o across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p015_20.png]
Figure 21
Figure 21. Figure 21: The mean of diagnosis ratings for Grok-4 across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p015_21.png]
Figure 22
Figure 22. Figure 22: The mean of diagnosis ratings for Deepseek-R1 across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p015_22.png]
Figure 23
Figure 23. Figure 23: The mean of diagnosis ratings for Llama across the stories in the dataset. Left: C [PITH_FULL_IMAGE:figures/full_fig_p016_23.png]
Figure 24
Figure 24. Figure 24: Abduction-guided LLM transformation successfully shifts narrative from C [PITH_FULL_IMAGE:figures/full_fig_p016_24.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 2 linked inside Pith

  1. [1]

    Penguin,

    [Achebe, 2012] Chinua Achebe.Hopes and impediments: Selected essays. Penguin,

  2. [5]

    Yale University Press,

    [Bercovitch, 1975] Sacvan Bercovitch.The Puritan origins of the American self. Yale University Press,

  3. [8]

    Conflictlens: Llm-based conflict resolution training in romantic relationship

    [Chunet al., 2025 ] Jiwon Chun, Gefei Zhang, and Meng Xia. Conflictlens: Llm-based conflict resolution training in romantic relationship. InAdjunct Proceedings of the 38th Annual ACM Symposium on User Interface Software and Technology, pages 1–3,

  4. [9]

    Green and Markus Ap- pel

    [Green and Appel, 2024] Melanie C. Green and Markus Ap- pel. Chapter one - narrative transportation: How stories shape how we see ourselves and the world. volume 70 of Advances in Experimental Social Psychology, pages 1–82. Academic Press,

  5. [13]

    Using large language models for narrative analysis: a novel application of gen- erative ai.Methods in Psychology, 12:100183,

    [Jenneret al., 2025 ] Sarah Jenner, Dimitris Raidos, Emma Anderson, Stella Fleetwood, Ben Ainsworth, Kerry Fox, Jana Kreppner, and Mary Barker. Using large language models for narrative analysis: a novel application of gen- erative ai.Methods in Psychology, 12:100183,

  6. [14]

    Subrahmanian

    [Kifer and Subrahmanian, 1992] Michael Kifer and V .S. Subrahmanian. Theory of generalized annotated logic pro- gramming and its applications**a preliminary report on this research has appeared in [34].The Journal of Logic Programming, 12(4):335–367,

  7. [17]

    [Markus and Kitayama, 2014] Hazel Rose Markus and Shi- nobu Kitayama

    Association for Computa- tional Linguistics. [Markus and Kitayama, 2014] Hazel Rose Markus and Shi- nobu Kitayama. Culture and the self: Implications for cognition, emotion, and motivation. InCollege student de- velopment and academic life, pages 264–293. Routledge,

  8. [21]

    [Mukherjeeet al., 2025 ] Sourabrata Mukherjee, Atul Kr Ojha, John Philip McCrae, and Ond ˇrej Du ˇsek. Evaluat- ing text style transfer evaluation: Are there any reliable metrics? InProceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technolo- gies (Volume 4: Student Re...

Show all 37 references
  1. [22]

    Generative agents: Interactive sim- ulacra of human behavior

    [Parket al., 2023 ] Joon Sung Park, Joseph O’Brien, Car- rie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. Generative agents: Interactive sim- ulacra of human behavior. InProceedings of the 36th an- nual acm symposium on user interface software and tec...

  2. [23]

    firehose of falsehood

    [Paul and Matthews, 2016] Christopher Paul and Miriam Matthews. The russian “firehose of falsehood” propaganda model.Rand Corporation, 2(7):1–10,

  3. [25]

    Using large language models for understanding narrative discourse

    [Piper and Bagga, 2024] Andrew Piper and Sunyam Bagga. Using large language models for understanding narrative discourse. In Yash Kumar Lal, Elizabeth Clark, Mohit Iyyer, Snigdha Chaturvedi, Anneliese Brei, Faeze Brah- man, and Khyathi Raghavi Chandu, editors,Proceedings of th...

  4. [26]

    [Reifet al., 2022 ] Emily Reif, Daphne Ippolito, Ann Yuan, Andy Coenen, Chris Callison-Burch, and Jason Wei

    Associa- tion for Computational Linguistics. [Reifet al., 2022 ] Emily Reif, Daphne Ippolito, Ann Yuan, Andy Coenen, Chris Callison-Burch, and Jason Wei. A recipe for arbitrary text style transfer with large language models. InProceedings of the 60th Annual Meeting of the Asso...

  5. [28]

    Cultural value alignment in large language models: A prompt-based analysis of schwartz values in gemini, chatgpt, and deepseek

    [Segerer, 2025] Robin Segerer. Cultural value alignment in large language models: A prompt-based analysis of schwartz values in gemini, chatgpt, and deepseek. 05

  6. [29]

    [Shakarian and Simari, 2022] Paulo Shakarian and Ger- ardo I. Simari. Extensions to Generalized Annotated Logic and an Equivalent Neural Architecture . In2022 Fourth International Conference on Transdisciplinary AI (TransAI), pages 63–70, Los Alamitos, CA, USA, Septem- ber

  7. [30]

    [Singelis and Brown, 1995] Theodore M Singelis and William J Brown

    IEEE Computer Society. [Singelis and Brown, 1995] Theodore M Singelis and William J Brown. Culture, self, and collectivist commu- nication: Linking culture to individual behavior.Human communication research, 21(3):354–389,

  8. [33]

    [Suzgunet al., 2022 ] Mirac Suzgun, Luke Melas-Kyriazi, and Dan Jurafsky

    Association for Compu- tational Linguistics. [Suzgunet al., 2022 ] Mirac Suzgun, Luke Melas-Kyriazi, and Dan Jurafsky. Prompt-and-rerank: A method for zero- shot and few-shot arbitrary textual style transfer with small language models.arXiv preprint arXiv:2205.11503,

  9. [34]

    Cat-llm: Style-enhanced large lan- guage models with text style definition for chinese article- style transfer.ACM Transactions on Knowledge Discovery from Data, 19(7):1–33,

    [Taoet al., 2025 ] Zhen Tao, Dinghao Xi, Zhiyu Li, Liumin Tang, and Wei Xu. Cat-llm: Style-enhanced large lan- guage models with text style definition for chinese article- style transfer.ACM Transactions on Knowledge Discovery from Data, 19(7):1–33,

  10. [35]

    [Tianet al., 2024 ] Yufei Tian, Tenghao Huang, Miri Liu, Derek Jiang, Alexander Spangher, Muhao Chen, Jonathan May, and Nanyun Peng. Are large language models capa- ble of generating human-level narratives? InProceedings of the 2024 Conference on Empirical Methods in Natural L...

  11. [36]

    Individualism and collec- tivism west view press.Boulder, CO,

    [Triandis, 1995] HC Triandis. Individualism and collec- tivism west view press.Boulder, CO,

  12. [37]

    [Yueet al., 2025 ] Matthew Yue, Zhikun Xu, Vivek Gupta, Thao Ha, Liesal Sharabi, and Ben Zhou. Relate- sim: Leveraging turning point theory and llm agents to predict and understand long-term relationship dynamics through interactive narrative simulations.arXiv preprint arXiv:2...

  13. [1975]

    The narrative construction of reality.Critical inquiry, 18(1):1–21,

    [Bruner, 1991] Jerome Bruner. The narrative construction of reality.Critical inquiry, 18(1):1–21,

  14. [1984]

    Can stories help llms rea- son? curating information space through narrative

    [Javadiet al., 2025 ] Vahid Sadiri Javadi, Johanne R Trippas, Yash Kumar Lal, and Lucie Flek. Can stories help llms rea- son? curating information space through narrative. InThe 2nd Workshop on Analogical Abstraction in Cognition, Perception, and Language (Analogy-Angle II), page 92,

  15. [1991]

    Individualism–collectivism and intergroup bias.Asian Journal of Social Psychology, 28(2):e70010,

    [Choi, 2025] Hoon-Seok Choi. Individualism–collectivism and intergroup bias.Asian Journal of Social Psychology, 28(2):e70010,

  16. [1992]

    Univ of South Carolina Press,

    [Klinkowitz, 2004] Jerome Klinkowitz.The Vonnegut Effect. Univ of South Carolina Press,

  17. [1993]

    Duke University Press,

    [Moon, 2020] Michael Moon.A small boy and others: Im- itation and initiation in American culture from Henry James to Andy Warhol. Duke University Press,

  18. [1994]

    Can LLMs narrate tabular data? an eval- uation framework for natural language representations of text-to-SQL system outputs

    [Singhet al., 2025 ] Jyotika Singh, Weiyi Sun, Amit Agar- wal, Viji Krishnamurthy, Yassine Benajiba, Sujith Ravi, and Dan Roth. Can LLMs narrate tabular data? an eval- uation framework for natural language representations of text-to-SQL system outputs. In Saloni Potdar, Lina R...

  19. [1995]

    The measurement of independent and interdependent self-construals.Personal- ity and social psychology bulletin, 20(5):580–591,

    [Singelis, 1994] Theodore M Singelis. The measurement of independent and interdependent self-construals.Personal- ity and social psychology bulletin, 20(5):580–591,

  20. [2000]

    [Hofstede, 1984] Geert Hofstede.Culture’s consequences: International differences in work-related values, volume

  21. [2004]

    In search of the long-tail: Systematic generation of long-tail inferen- tial knowledge via logical rule guided search

    [Liet al., 2024 ] Huihan Li, Yuting Ning, Zeyi Liao, Siyuan Wang, Xiang Lorraine Li, Ximing Lu, Wenting Zhao, Faeze Brahman, Yejin Choi, and Xiang Ren. In search of the long-tail: Systematic generation of long-tail inferen- tial knowledge via logical rule guided search. In Yas...

  22. [2012]

    Pyreason: Software for open world temporal logic,

    [Adityaet al., 2023 ] Dyuman Aditya, Kaustuv Mukherji, Srikar Balasubramanian, Abhiraj Chaudhary, and Paulo Shakarian. Pyreason: Software for open world temporal logic,

  23. [2014]

    Guilford press,

    [McAdams, 1993] Dan P McAdams.The stories we live by: Personal myths and the making of the self. Guilford press,

  24. [2016]

    Swag: Storytelling with action guid- ance

    [Peiet al., 2024 ] Jonathan Pei, Zeeshan Patel, Karim El- Refai, and Tianle Li. Swag: Storytelling with action guid- ance. InFindings of the Association for Computational Linguistics: EMNLP 2024, pages 14086–14106,

  25. [2020]

    Are large language models actu- ally good at text style transfer?CoRR,

    [Mukherjeeet al., 2024 ] Sourabrata Mukherjee, Atul Kr Ojha, and Ondrej Dusek. Are large language models actu- ally good at text style transfer?CoRR,

  26. [2022]

    Biased tales: Cultural and topic bias in generating children’s stories

    [Rooeinet al., 2025 ] Donya Rooein, Vil ´em Zouhar, Debora Nozza, and Dirk Hovy. Biased tales: Cultural and topic bias in generating children’s stories. InProceedings of the 2025 Conference on Empirical Methods in Natural Lan- guage Processing, pages 52–72,

  27. [2023]

    The living tradition.Gen- eral history of Africa, 1:166–205,

    [Bˆa, 1981] Amadou Hampat´e Bˆa. The living tradition.Gen- eral history of Africa, 1:166–205,

  28. [2024]

    The role of transportation in the persuasiveness of public narratives.Journal of personality and social psy- chology, 79(5):701,

    [Green and Brock, 2000] Melanie C Green and Timothy C Brock. The role of transportation in the persuasiveness of public narratives.Journal of personality and social psy- chology, 79(5):701,

  29. [2025]

    Sea-cret agents: Mar- itime abduction for region generation to expose dark vessel trajectories.ArXiv, abs/2502.01503,

    [Bavikadiet al., 2025b ] Divyagna Bavikadi, Nathaniel Lee, Paulo Shakarian, and Chad Parvis. Sea-cret agents: Mar- itime abduction for region generation to expose dark vessel trajectories.ArXiv, abs/2502.01503,

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.