Pith. sign in

REVIEW 3 major objections 5 minor 30 references

CoMet: Metaphor-Driven Covert Communication for Multi-Agent Language Games

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read CoMet gives LLM agents metaphor reasoning and self-improving generation, raising win rates in two covert-communication games.

desk verdict Good idea, shaky evidence: the self-improvement claim is undercut by experience-pool contamination on the same word pairs used for evaluation. read the letter →

arxiv 2505.18218 v1 pith:CYH2GCN2 submitted 2025-05-23 cs.CL cs.AI

Shuhang Xu , Fangwei Zhong This is my paper · ORCID
classification cs.CLcs.AI
keywords metaphorreasoningcovertcommunicationmulti-agentlanguagegameslargemodelsUndercoverAdversarialTabooself-improvinggenerationsemanticevasion
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 sets out to show that metaphors can be a working communication strategy for LLM agents, not just a literary decoration. It introduces CoMet, a framework that adds a hypothesis-testing metaphor reasoner and a self-improving metaphor generator to a standard LLM game agent, and evaluates it in two multiplayer language games where concealment and misdirection decide the outcome. The central empirical claim is that CoMet materially improves covert communication: in Adversarial Taboo, attacker win rate rises 47% and defender win rate 30% over baselines, and in Undercover, undercover agents' win rate increases from 0.20 with Chain-of-Thought to 0.35 with CoMet. If true, the result matters because covert communication—saying something that allies decode but opponents cannot—is a general social capability, and language games stress the exact skills an agent would need for it.

What carries the argument

The carrying object is a five-stage agent pipeline with two metaphor-specific components. The Hypothesis-Based Metaphor Reasoner receives any utterance that does not literally fit the agent's secret word, frames two competing hypotheses (speaker means my word, or speaker means the other word), expands the utterance into ontological, structural, and spatial metaphor aspects, and scores the match between those aspects and the word's features on a six-level scale before accepting or rejecting the hypothesis. The Metaphor Generator maintains a self-play experience pool: each generated metaphor is stored with its explanation, evaluator feedback, usage statistics, and success score, and after every five games low-performing, over-referenced entries are pruned while new high-quality ones replace them, so the agent's metaphorical speech improves with accumulated gameplay.

What would settle it

Run the same two games with human annotators labeling each speech for whether it leaked the secret word and each vote for whether it identified the correct teammate, over enough episodes to compute confidence intervals; CoMet's claimed advantage would be refuted if its win-rate gap over baselines disappears or the human labels contradict the LLM-judged metrics.

Watch

Extended reading notes

Core claim

CoMet's central claim is that LLM agents fail at covert communication mainly because they interpret and produce language literally, and that a scaffolded metaphor process fixes this. The paper argues that in Undercover, civilians can use metaphors as 'concept camouflage': a clue like 'it wears a crown' lets teammates identify 'pineapple' while the undercover team cannot pin down the word. In Adversarial Taboo, the same machinery lets an attacker steer a defender toward a secret word or a defender avoid it, by substituting metaphorical descriptions for direct ones. The authors report that the framework generalizes across six LLMs and that each component contributes, with the Self-Monitor module being especially load-bearing: removing it drops the Undercover win rate to 0.05, below the Chain-of-Thought baseline.

Load-bearing premise

The central claim assumes that the win rates and the LLM-computed metrics of privacy protection, feature extraction, identity assessment, and deceptive statements faithfully measure real covert-communication ability, despite having no human annotation or significance testing.

Editorial extensions

If this is right

  • In Adversarial Taboo, CoMet outperforms baseline agents as both attacker and defender, with reported win-rate gains of 47% and 30% respectively.
  • In Undercover, CoMet raises the undercover win rate from 0.20 (CoT) to 0.35 and the civilian win rate to 0.85 when facing the CoT baseline.
  • The win-rate advantage generalizes across the models tested, and CoMet reduces failure rates below 15% for every LLM in the Adversarial Taboo generalization experiment.
  • The ablation results imply that identity self-monitoring, not metaphor alone, is necessary for the gains: without it, CoMet's win rate drops to 0.05.

Reading between the lines

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

  • Because the reasoner deliberately reduces metaphor interpretation to a binary match against a secret word, the same design could be lifted out of games into any setting where two parties share private context and want to exclude eavesdroppers, such as identity verification or protected coordination.
  • A natural stress test is to push the experience pool beyond its 100-entry cap and beyond the food and animal word domains; the paper's own statistics suggest metaphor categories differ in quality, so domain transfer is not guaranteed.
  • The quantitative metrics are computed by LLM judgment on game logs with no reported human annotation or significance testing, so a human-labeled replication is the most direct way to see whether the covert-communication advantage is real or partly an artifact of the evaluator model.
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

3 major / 5 minor

Summary. CoMet is a framework that augments LLM-based agents with a hypothesis-based metaphor reasoner and a self-improving metaphor generator, and evaluates it in two multi-agent language games: Undercover and Adversarial Taboo. The paper reports that CoMet raises attackers' win rates by 47% and defenders' by 30% in Adversarial Taboo compared with baselines, and that in Undercover the undercover win rate rises from 0.20 with CoT to 0.35 with CoMet (Table 1). Additional metrics (FER, OIAA, SIAA, PPC, IISC) computed from game logs are used to argue that CoMet improves feature extraction, identity reasoning, privacy protection, and deceptive capability. The central claim is that metaphor-driven reasoning and generation significantly improve covert and deceptive communication in these language games.

Significance. The paper opens a genuinely underexplored direction: using metaphors as a communication strategy in interactive multi-agent language games rather than on static metaphor-detection or generation benchmarks. Its concrete assets include an implemented framework, released code, a curated dataset of 200 word pairs, experiments across five different LLMs, and an ablation study of the main components. If the empirical claims hold, the work would be a useful demonstration of LLM agents exploiting metaphorical expression for concealment, misdirection, and privacy-aware communication. However, the current evaluation does not establish that the improvements generalize beyond the exact word pairs used for experience accumulation, and the reported magnitudes are not supported by error bars, confidence intervals, or human-validated metrics. The significance is therefore conditional on resolving these evaluation threats.

major comments (3)
  1. [Appendix C, Section 4.1, Figure 6(b)] The experience pool stores exact word-pair identities, as shown in Figure 7, which records the entry "words": ["snake","lizard"], and Section 4.1 states that the 200 word pairs used for evaluation are each tested across 10 episodes. Because the metaphor generator retrieves and reuses experiences across games, including self-play on the same evaluation pairs, the improvements reported in Figure 6(b) and Table 1 may reflect pair-specific memorization rather than a generalizable metaphor-generation capability. The paper provides no held-out word-pair split and no comparison on unseen pairs, so the "self-improving" claim is not yet established. Please add a held-out evaluation in which the experience pool is trained on one set of word pairs and evaluated on a disjoint set, or otherwise demonstrate that gains transfer to pairs never encountered during experience accumulation.
  2. [Section 4.1, Table 4, Appendix D] The metrics FER, OIAA, SIAA, PPC, and IISC (defined in Table 4 and Appendix D) are computed by LLM-based analysis of game logs, without independent human annotation or grounding in a held-out benchmark. Since these metrics support the non-win-rate claims in Table 1 and Figure 5, their reliability is load-bearing for the paper's assertions about feature extraction, identity reasoning, and privacy protection. The authors should provide human validation on a sample of logs, report inter-annotator agreement, or otherwise demonstrate that the LLM judgments correspond to verifiable ground truth.
  3. [Section 4.2, Table 1, Figure 4] No variability measure or statistical test is reported for any of the main results. Figure 4 states attackers' win rates increased by 47% and defenders' by 30%, and Table 1 reports differences such as 0.35 vs 0.20, but there are no confidence intervals, per-pair distributions, or significance tests. With only 200 word pairs and 10 episodes per pair, the reader cannot assess whether these differences are robust or within the noise of the games. Please report per-pair win-rate distributions, bootstrap confidence intervals, or appropriate significance tests for the key comparisons.
minor comments (5)
  1. [Table 2] The column header "Win Rate↓" appears to use the wrong arrow, because the full CoMet row has the highest value (0.70) and the paper presents it as the best configuration; presumably ↑ was intended.
  2. [Figure 4(b)] The axis labels in Figure 4(b) are difficult to parse; please clarify whether each entry denotes attacker-method versus defender-method, and specify which role is fixed as CoT when reading the curves.
  3. [Appendix F, Figure 19] Adversarial Taboo appears to be run with a simplified prompt that does not instantiate the full CoMet pipeline shown in Figure 2 (Feature Extractor, Belief Mapper, Self-Monitor, Strategy Planner). Please clarify which modules are active in that game and whether the reported CoMet results use the same framework as Undercover.
  4. [Section 4.1, Balanced Metrics] The Balanced Metrics formula M'_i = avg(M_i^Civ, M_i^Und) - Var(M_i^Civ, M_i^Und) is nonstandard and can produce values outside the range of the original metrics; please provide a justification for subtracting the variance or use a more conventional correction.
  5. [Ethical Statement] The assertion that "it is infeasible to use our method to make LLMs output risky content" is unsupported by the experiments, which only test daily-vocabulary game words; please soften or remove this categorical claim.

Circularity Check

1 steps flagged · score 6.0 of 10

CoMet's experience pool records the exact test word pairs and is populated on the same 200 pairs later used for evaluation, so the reported self-improvement gains may reduce to pair-specific memorization rather than generalizable metaphor skill.

  1. fitted input called prediction [Section 4.1 Experimental Setups; Section 4.3 / Figure 6(b); Appendix C 'Experience Pool Structure and Maintenance'; Figure 7]
    "We collected 200 word pairs ... and each pair is tested across 10 evaluation episodes. ... The experience pool for metaphor generation is structured as a dictionary format collection ... Each experience entry contains: Text content, Labels, Statistics. ... Figure 7 stores 'words': ['snake','lizard'], 'metaphor': 'They are silent dancers.', 'teammate_recognitions': 7, 'rival_recognitions': 1. ..."

    The experience pool stores the exact word pair being evaluated, including the rival word, and persists across games with dynamic retrieval 'based on scores and metaphor categories'. Because no held-out word pairs are reported, the same 200 pairs that populate the pool are also the pairs on which Figure 6(b) and Table 1 measure success. At 50/100 experiences, the generator can retrieve a metaphor already fitted to the exact pair under test from earlier self-play episodes, so the reported improvement is not independent evidence of generalizable metaphor reasoning or generation; it is partly retrieval from a per-pair memory that was fitted on the evaluation data. The claimed 'self-improvement' is thus a fitted input being reported as a prediction.

full rationale

Most of CoMet is not definitionally circular: the Feature Extractor, Belief Mapper, Strategy Planner, and the hypothesis-testing reasoner are described as prompted LLM components with a stated scoring procedure, and the headline win rates come from actual game terminations. However, the paper's distinctive contribution, the self-improving metaphor generator, is evaluated on the same 200 word pairs used to populate its experience pool, whose entries record the exact word pair, the rival word, and per-pair recognition counts. Since retrieval selects 'relevant experiences' by score and category, the 0/50/100-experience success curve and the CoMet win rates in Table 1 can be driven by retrieving cached per-pair metaphors rather than by improved metaphor skill. No held-out word-pair evaluation is reported. A secondary self-referentiality concern is that the experience-pool feedback and several reported metrics (FER, PPC, IISC, and the Figure 6(b) success rate) are based on LLM log-reading rather than independent human annotation; this weakens the evidence further but is not needed for the main reduction. Overall, the central claim is partially circular because the self-improvement signal is fitted to the evaluation instances and then presented as a prediction.

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

The central claim rests on several domain assumptions: that the injected metaphor taxonomy helps, that binary hypothesis testing captures metaphor interpretation in these games, that LLM-generated logs can be scored without human validation, and that the two benchmark games are representative. The framework introduces no new physical or conceptual entities, but Algorithm 1's threshold, weights, and scoring rubric are hand-set and undisclosed, and these directly control the reasoner's decisions.

free parameters (4)
  • Hypothesis-testing score threshold T = Not reported
    Algorithm 1 uses a threshold T on the semantic matching score to decide whether a metaphor refers to the agent's secret word; the value is not disclosed, so the binary classification outcome depends on an unstated hyperparameter.
  • Position-based weight factors w_f and w_m = Not reported
    Algorithm 1 applies position-based coefficients to features and metaphorical aspects, giving priority to first-identified characteristics; the values are hand-designed and not reported.
  • Semantic matching function delta with discrete scores = Scores in {0, 0.2, 0.4, 0.6, 0.8, 1.0}
    The mapping from feature-metaphor coherence to these six discrete scores is not specified and is effectively a hand-set scoring rubric that directly controls the reasoner's decisions.
  • Experience pool capacity and pruning thresholds = Max 100 experiences per category; prune after 5 games for references over 5 with low score
    The self-improving metaphor generator depends on these capacities and cutoffs, which are chosen by the authors rather than derived from data.
assumptions (5)
  • domain assumption Conceptual metaphor theory categories of ontological, structural, and spatial metaphors are sufficient and useful for guiding LLM metaphor reasoning and generation.
    Section 3.3 and Appendix F inject this theory as prior knowledge; if the taxonomy does not match the game's communicative demands, the framework's mechanism is weakened.
  • domain assumption Descriptions that do not align with the current game focus are likely metaphors, and metaphor understanding can be reduced to a binary hypothesis about whether the utterance refers to the agent's own word.
    Section 3.3, Hypothesis-Based Metaphor Reasoner; this simplification is core to the reasoner's operation.
  • domain assumption LLM game logs and LLM-based evaluation reliably reflect feature extraction, identity assessment, privacy leakage, and deceptive statements.
    Section 4.1 and Appendix D define metrics on game logs without human annotation or external benchmark validation.
  • domain assumption Outcomes in Undercover and Adversarial Taboo transfer to broader covert-communication settings such as negotiation, privacy, and alignment.
    The paper's stated motivation includes these broader settings, but only two language games are evaluated.
  • domain assumption Self-play experience accumulated in one game generalizes to later games and to different LLM backbones.
    Section 3.5 describes an experience pool that is reused across games; no controlled test of cross-game transfer is reported, and generalization across LLMs is tested only within the same two games.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CoMet: Metaphor-Driven Covert Communication for Multi-Agent Language Games." pith.science (2026). https://pith.science/paper/CYH2GCN2

@misc{pith2026250518218,
  author       = {Pith},
  title        = {Pith review of: CoMet: Metaphor-Driven Covert Communication for Multi-Agent Language Games},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CYH2GCN2}},
  note         = {Machine review of arXiv:2505.18218}
}
read the original abstract

Metaphors are a crucial way for humans to express complex or subtle ideas by comparing one concept to another, often from a different domain. However, many large language models (LLMs) struggle to interpret and apply metaphors in multi-agent language games, hindering their ability to engage in covert communication and semantic evasion, which are crucial for strategic communication. To address this challenge, we introduce CoMet, a framework that enables LLM-based agents to engage in metaphor processing. CoMet combines a hypothesis-based metaphor reasoner with a metaphor generator that improves through self-reflection and knowledge integration. This enhances the agents' ability to interpret and apply metaphors, improving the strategic and nuanced quality of their interactions. We evaluate CoMet on two multi-agent language games - Undercover and Adversarial Taboo - which emphasize Covert Communication and Semantic Evasion. Experimental results demonstrate that CoMet significantly enhances the agents' ability to communicate strategically using metaphors.

Figures

Figures reproduced from arXiv: 2505.18218 by the authors.

Figure 1
Figure 1. Comparison of three communication strategies—Straightforward Description, Concealment, and Metaphor [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of the CoMet framework, illustrated within the “concept camouflage” task in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The metaphor reasoning process based on hypothesis testing when players holding the word “kite” encounter the statement “homesick bird.” The process involves hypothesizing whether the metaphor refers to a kite (H0) or another object (H1), followed by analysis of features such as flight, lifelessness, and being tethered. Through metaphor expansion and hypothesis testing, the model determines that the metaphor best fi… view at source ↗
Figures from the paper (17 more)
Figure 4
Figure 4. Figure 4: Performance comparison of different LLMs [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 6
Figure 6. Figure 6: Performance comparison of different methods [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: An example demonstrating the structure of data stored in the experience pool. [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: A case and explanation for undercovers’ self-protection and deception. Team with “Butterfly” uses CoMet [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: A case of civilians’ metaphorical communication. The team with “Wolf” uses CoMet and the team with [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: Full game logs of the case mentioned by Figure [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]
Figure 11
Figure 11. Figure 11: The prompt for Feature Extractor # Background At the beginning of the game, each player is randomly assigned one of two words that share some similarities. Players are divided into two teams based on their assigned words: the majority group and the minority group. The…
Figure 12
Figure 12. Figure 12: The prompt for Belief Mapper [PITH_FULL_IMAGE:figures/full_fig_p019_12.png]
Figure 13
Figure 13. Figure 13: The prompt for Self-Monitor # Background At the beginning of the game, each player is randomly assigned one of two words that share some similarities. Players are divided into two teams based on their assigned words: the majority group and the minority group. The team…
Figure 14
Figure 14. Figure 14: The prompt for Strategy Planner [PITH_FULL_IMAGE:figures/full_fig_p020_14.png]
Figure 15
Figure 15. Figure 15: The prompt for Actor (Speaker) # Background At the beginning of the game, each player is randomly assigned one of two words that share some similarities. Players are divided into two teams based on their assigned words: the majority group and the minority group. The t…
Figure 16
Figure 16. Figure 16: The prompt for Actor (Voter) [PITH_FULL_IMAGE:figures/full_fig_p021_16.png]
Figure 17
Figure 17. Figure 17: The prompt for metaphor generation step 1. [PITH_FULL_IMAGE:figures/full_fig_p022_17.png]
Figure 18
Figure 18. Figure 18: The prompt for metaphor generation step 2. [PITH_FULL_IMAGE:figures/full_fig_p023_18.png]
Figure 19
Figure 19. Figure 19: The prompt for Adversarial Taboo (Simplified) [PITH_FULL_IMAGE:figures/full_fig_p024_19.png]
Figure 20
Figure 20. Figure 20: The collection of 100 animal-themed word pairs for Undercover [PITH_FULL_IMAGE:figures/full_fig_p025_20.png]
Figure 21
Figure 21. Figure 21: The collection of 100 food-themed word pairs for Undercover [PITH_FULL_IMAGE:figures/full_fig_p026_21.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 27 canonical work pages

  1. [1]

    **Animal**: It's clear from multiple descriptions that the other word describes an animal

    work page

  2. [2]

    Ekaterina Shutova

    Reflexion: an autonomous agent with dy- namic memory and self-reflection.ArXiv Preprint ArXiv:2303.11366. Ekaterina Shutova. 2010. Models of metaphor in nlp. In Proceedings of the 48th Annual Meeting of the Associ- ation for Computational Linguistics, pages 688–697. Paul H. Thibodeau, Teenie Matlock, and Stephen J. Flus- berg. 2019. The role of metaphor i...

    arXiv 2010

  3. [3]

    butterfly

    Cultural and Contextual Awareness: Consider cultural associations and contexts to strengthen metaphors. } Figure 7: An example demonstrating the structure of data stored in the experience pool. following a forward reasoning path from literal to metaphorical meaning, leveraging prior knowledge to reduce the breadth of metaphor reasoning. For in- stance, du...

    work page 2024

  4. [4]

    Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan

    Language agents with reinforcement learn- ing for strategic play in the werewolf game.ArXiv Preprint ArXiv:2310.18940. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan

    arXiv

  5. [5]

    asymmetric encryption

    Tree of thoughts: Deliberate problem solving with large language models. InAdvances in Neural Information Processing Systems, volume 36, pages 11809–11822. Curran Associates, Inc. Yuan Yao, Haoxi Zhong, Zhengyan Zhang, Xu Han, Xiaozhi Wang, Kai Zhang, Chaojun Xiao, Guoyang Zeng, Zhiyuan Liu, and Maosong Sun. 2021. Adver- sarial language games for advanced...

    arXiv 2021

  6. [6]

    Clarity and Simplicity: Use clear and simple metaphoric language to invoke strong imagery

    work page

  7. [9]

    "pack behavior,

    **Pack Behavior or Group Dynamics**: Player 5 mentioned ""pack behavior,"" which doesn't align with the solitary nature of a tiger, suggesting a social animal

    work page

  8. [10]

    ### Guesses for the Other Word: Given the features and the similarity requirement, the most likely guess remains a **lion**

    **Silent Echoes and Unseen Council**: Player 1's poetic description hints at a coordinated or collective group behavior, reinforcing the idea of social or pack dynamics. ### Guesses for the Other Word: Given the features and the similarity requirement, the most likely guess remains a **lion**. Lions are known for their pride and pack behavior, which align...

    work page

Show all 30 references
  1. [11]

    "In the realm of silent echoes, the unseen council navigates the wilderness

    **Identifying Metaphors:** - The phrase ""In the realm of silent echoes, the unseen council navigates the wilderness"" is likely a metaphor. It uses poetic language that isn't directly descriptive of a physical characteristic

  2. [12]

    "wolf"" and the guessed word

    **Understanding the Metaphor:** - Now, let's compare this metaphor to the features of ""wolf"" and the guessed word ""tiger."" - ""Silent echoes"" may imply the howl of wolves, and ""unseen council"" could imply stealth and a coordinated group, which aligns with the behavior o...

  3. [13]

    "wolf,"" it's reasonable to infer that Player 1 is likely describing the word

    **Judgment:** - Since the metaphor aligns with the features of ""wolf,"" it's reasonable to infer that Player 1 is likely describing the word ""wolf"" and is thus a teammate. Based on this analysis, Player 1's metaphorical statement seems to come from a teammate rather than an...

  4. [14]

    Find those descriptions that not match your word, then they should be describing the other word

    Check if other players' description aligns with your word. Find those descriptions that not match your word, then they should be describing the other word

  5. [15]

    - These features should not fit your word, but rather fit only the other word

    Analyze those suspicious descriptions, extract and summarize the unique features of the other word. - These features should not fit your word, but rather fit only the other word. They are the key information you use to identify the other word

  6. [16]

    You are player {id}, and you were assigned a word '{word}'

    Generate or adjust your guess for the other word: - The guess should be based on two pieces of information: one is the features you just summarized, and the other is the basic principle that the other word is similar to or related to your word. - If it is the first time genera...

  7. [17]

    Player with 'Word A': ...; Player with 'Word B': ...; Uncertain:

    group the players who describe the unique features of the same word into two teams: - For those who describe the unique characteristics of a word, you can assume that they are assigned to this word (your word or your guessed word); - For those whose descriptions can both match...

  8. [18]

    You are player {id}, and you were assigned a word '{word}'

    try to refer to the voting results to determine the uncertain players and assign them to a team: - if his voting object is always Team A, then he may belong to Team B - If his voting result is same with many players, then he may belong th the team with those players. # Informa...

  9. [19]

    Review your strategy and the guesses about the other word, understand the requirements of strategy for your action

  10. [20]

    Based on the requirements that the features need to meet, list some candidate features to describe

  11. [21]

    If all candidate features are used, try to give some new candidates and retry to choose one

    Refer to the game history and select one candidate feature that no other player has described. If all candidate features are used, try to give some new candidates and retry to choose one

  12. [22]

    It has <selected feature>

    Describe the selected feature in brief, without the need for additional explanation or exposure of other features. Be careful not to directly expose words, but to use pronouns instead. Example: "It has <selected feature>.", "this thing is <selected feature>.", etc. # Informati...

  13. [23]

    Review your definition of your own identity

  14. [24]

    Figure out your winning conditions based on your identity

  15. [25]

    You are player {id}, and you were assigned a word '{word}'

    Choose one player to vote. # Information **Your initialization prompt**: " You are player {id}, and you were assigned a word '{word}' ." **Your analysis of another word's features** \n"{feature}"\n **Your analysis of your camp and identity** \n"{identity}"\n **history of other...

  16. [26]

    things" or

    Understand three types of metaphor, namely: - ONTOLOGICAL METAPHOR: Ontological metaphors are those in which abstract concepts or experiences are understood as having an existence or being in some form of objector substance. This metaphor involves treating abstract concepts li...

  17. [27]

    You will get more information about this method

    Analyze the features you have listed and identify the most suitable one feature for generating metaphors to achieve the goal of conveyinginformation to teammates rather than opponents, as well as the appropriate method for generating metaphors. You will get more information ab...

  18. [28]

    Review the informations that you need: - What's your word to describe? - What features do you choose to describe in words, or what things are associated with the expansion of words? - What's the method you choose to generate metaphor?\n

  19. [29]

    And your opponent cannot accurately locate a feature from your metaphor and infer words from the feature, so they cannot guess your words

    Your teammates will try to understand your metaphor by comparing the features of each word with your description. And your opponent cannot accurately locate a feature from your metaphor and infer words from the feature, so they cannot guess your words. Please aim for this effect

  20. [30]

    {method_}

    Refer to the theory and case studies of this metaphorical approach, generate your own metaphorical description (a concise sentence) of your word. - The theory is in the following format: "{method_}" - {use1}\n "{metaphor1}"\n This metaphor is describing {word11}. {explain1} {c...

  21. [2023]

    InProceedings of the 2023 Con- ference on Game-based AI

    Avalonbench: Evaluating llms playing the game of avalon. InProceedings of the 2023 Con- ference on Game-based AI. Details about the exact conference are missing. Yujie Lin, Jingyao Liu, Yan Gao, Ante Wang, and Jin- song Su. 2024. A dual-perspective metaphor detec- tion framewo...

    2023 arXiv

  22. [2024]

    InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 7315–7332

    Magic: Investigation of large language model powered multi-agent in cognition, adaptability, ra- tionality and collaboration. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 7315–7332. Zelai Xu, Chao Yu, Fei Fang, Yu Wang, and Yi Wu

    2024

Pith tools

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