Aligning Large Language Models for Faithful Integrity Against Opposing Argument

See-kiong Ng; Tat-Seng Chua; Yang Deng; Yong Zhao

REVIEW 5 major objections 5 minor 39 references

Aligning Large Language Models for Faithful Integrity Against Opposing Argument

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Aligned to its own confidence estimates, a language model defends correct answers against opposing arguments and concedes when its answer is wrong, raising final-answer accuracy across debate-style reasoning benchmarks.

desk verdict A plausibly useful self-confidence alignment recipe with a genuinely new confidence estimator, but the evaluation's out-of-distribution calibration gap and underspecified baselines keep it short of a clean accept. read the letter →

arxiv 2501.01336 v1 pith:UVXB45G5 submitted 2025-01-02 cs.CL

Yong Zhao , Yang Deng , See-Kiong Ng , Tat-Seng Chua This is my paper · ORCID

classification cs.CL

keywords faithfulintegrityopposingargumentconfidenceestimationDirectPreferenceOptimizationsemanticentropyLLMalignmentconversationalrobustnessbilateral

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

Large language models cave too easily when a user argues back, even when the original answer was right, and they dig in too hard when the user is right and they are wrong. The paper claims the two failure modes share one fix: align the model to its own confidence. The authors compute a two-part confidence score for each answer, build preference pairs that tell the model to persist when confident and yield when unsure, and fine-tune on those pairs. Across mathematics, logic, commonsense, and generic reasoning benchmarks, the final-answer accuracy after a two-round debate rises, from 0.598 to 0.742 in the generic category for LLaMA-3-8B. If the claim holds, conversational models can be made more reliable without retrieval, fact-checking, or external knowledge at inference time.

What carries the argument

Bilateral Confidence Estimation (BCE), a two-stage confidence score. A regressor trained on the model's internal hidden states (the 26th layer output at the last token) approximates semantic entropy, giving a question-level confidence; a per-answer adjustment multiplies it by a cumulative probability ratio computed from the length-normalized log-likelihood of the answer among a multinomial beam sample. Confidence thresholds at the 66.7th and 33.3rd percentiles map five hand-written response stances (persist, slight concession, neutral, lean opposing, fully agree) into positive and negative sets for Direct Preference Optimization. The same score decides both persistence and concession, which is what makes a single alignment pass address both failure modes.

What would settle it

On a subset of MMLU where the base model's bilateral confidence is high but the answer is wrong, present the correct answer as the user's counter-argument; if the AFICE-tuned model concedes no more often than it defends correct high-confidence answers, the confidence-as-truth-proxy assumption is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a model aligned to its own calibrated confidence will behave with faithful integrity in argument: it defends an originally correct statement against an incorrect user counter-argument and corrects an originally incorrect statement when the user's counter-argument is faithful. The evidence is that AFICE fine-tuning raises the average proportion of final answers matching the correct answer across all four question categories and both base models, with the confidence-based method beating verbalization, semantic entropy, P(True), and predictive entropy baselines. The paper presents the gain as coming from aligning response stance to confidence rather than from injecting correct content, since the preference data are built from the model's own answers and confidence scores.

Load-bearing premise

The load-bearing premise is that the model's self-estimated confidence tracks actual correctness well enough to serve as the training signal, so confidently wrong answers would be trained to be stubbornly defended and the paper does not analyze that case.

Editorial extensions

If this is right

A conversational LLM trained this way should show measurably fewer sycophantic U-turns when correct and more willingness to be corrected when wrong, without needing retrieval or fact-checking at inference time.
The cheap confidence estimate—using early-truncated tokens (K=60) and 20 samples—makes the alignment practical in interactive latency budgets where full-sequence semantic entropy is too slow.
Because one confidence score drives both persistence and concession, the method should transfer to any QA-style domain where internal states predict correctness, not just the four benchmark categories tested.
The consistent gain over P(True) and predictive entropy indicates that combining question-level internal-state certainty with answer-level likelihood rank is what lets the model decide when to yield.

Reading between the lines

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

The paper's own limitation note predicts that fine-tuning shifts internal states, so the confidence regressor may need refitting after alignment; a concrete next experiment is to measure AUROC of the regressor on the fine-tuned model and see whether periodic re-estimation restores performance.
Because the preference labels come from self-estimated confidence, the method inherits any systematic overconfidence; a testable extension is to run AFICE on a deliberately miscalibrated model and check whether the accuracy gains shrink or reverse.
The user counter-arguments are machine-generated as either fully correct or fully incorrect; real-user arguments are often partially valid, so a harder extension would test AFICE against mixed-validity arguments and finer-grained concession thresholds.
The confidence score could be extended from a two-way persist-or-concede decision to a three-way decision that includes abstention ('I don't know') or a retrieval call at very low confidence.

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.

Desk Editor's Note

read the letter

You should know two things about this paper. First, the central idea is more interesting than the average alignment patch: instead of training the model to match external correctness labels, AFICE trains it to match its own estimated confidence, using a bilateral confidence estimator that combines internal-state regression with cumulative probability ratios. That combination is new, and the preference construction (five response types, thresholds at confidence percentiles, six DPO pairs) is concrete and reproducible. Second, the main table shows real, reasonably consistent gains across two base models and several benchmarks, with especially strong improvements on the LLM-False scenario. That's a solid empirical result worth engaging with.

What the paper does well: BCE is efficient (early token truncation at K=60, N=20), and the comparison against Semantic Entropy, P(True), and Predictive Entropy is fair in spirit because all are used as confidence signals for the same alignment recipe. The appendix reporting LLM Correct vs. LLM False separately, plus Both/Either, is more informative than a single average. The limitations section is honest: C.1 explicitly flags that the regressor may need retraining after fine-tuning, which is exactly the soft spot the stress-test note hits.

Now the soft spots, in proportion. The black-box baseline protocol is genuinely underspecified: it's unclear how the verbalization or semantic entropy score is turned into a final response decision, so part of the advantage could come from how the policy is adjusted rather than the confidence estimator itself. No error bars or significance tests either, and some margins are thin (Vicuna AFICE 0.623 vs. Predictive Entropy 0.616 on GSM8K). The deeper concern is the one the paper's own limitation C.1 gestures at: the confidence regressor is validated only on MMLU, which is multiple-choice and in-distribution, while the deployment tasks are open-ended and out-of-distribution. If the base model is confidently wrong on those tasks, DPO will reinforce exactly the stubborn errors the paper is trying to fix. The high LLM-False accuracy (0.990 for LLaMA-3 AFICE) suggests the effect is limited in practice, but we don't get a calibration or accuracy-versus-confidence breakdown on the eval tasks, so the worry is not directly retired.

Who this is for: people working on LLM alignment, factuality, and conversational robustness. It deserves a serious referee, not a desk reject. The authors should be asked to clarify the black-box baseline protocol, add error bars or significance tests, and provide a calibration or accuracy-vs-confidence breakdown on at least one out-of-distribution eval task. If those come back clean, this could be a useful recipe.

Referee Report

5 major / 5 minor

Summary. The paper introduces AFICE, a framework that combines Bilateral Confidence Estimation (BCE) with Direct Preference Optimization (DPO) to make LLMs more robust to opposing arguments in two-round conversations. BCE first estimates question-level confidence by regressing internal hidden states onto semantic entropy computed on a subset of MMLU, then adjusts this estimate for a specific answer using a cumulative probability ratio over sampled responses. These confidence scores are used to label five possible response types (persist, slight concession, neutral, lean toward opposing view, full agreement) as preferred or dispreferred, and DPO fine-tunes the model to follow the labels. The method is evaluated on math (GSM8K), first-order logic (PrOntoQA), commonsense (StrategyQA, CSQA2, CREAK), and generic reasoning (nine BIG-Bench tasks), in two settings: when the model's initial answer is correct and the user argues incorrectly, and when the initial answer is incorrect and the user argues correctly. Averaged results over both settings are reported for Vicuna-7B and LLaMA-3-8B, with AFICE obtaining the highest average accuracy on most categories.

Significance. If the reported results are reliable, the paper makes a useful practical contribution by demonstrating that a model's own internal-state-based confidence can be used as a training signal to influence whether the model persists in or revises its initial answer during argumentative conversations. The proposed BCE method is computationally cheaper than full-sequence semantic entropy, and the paper provides extensive evaluation tables across several reasoning benchmarks. The conceptual idea of aligning conversational behavior with self-estimated confidence is interesting and could be applicable beyond the specific two-round setting. However, the significance of the work depends on whether the evaluation is fair and whether the confidence signal is actually predictive of answer correctness on the deployment tasks; both points need substantial additional support before the empirical claims can be accepted.

major comments (5)

[§4.1, Appendix B.2] The evaluation protocol for the black-box baselines (Verbalization and Semantic Entropy) is underspecified. The paper states that black-box methods are applied directly to the evaluation dataset and that the confidence score is used as an indication of the LLM's agreement with the initial answer, but it never explains how this score influences the final response in the two-round conversation. Without a precise mechanism — for example, a threshold above which the model is instructed to persist and below which it is instructed to yield — the accuracies reported for these baselines in Table 1 are not interpretable, and the comparison with AFICE, which uses a concrete DPO training procedure, may be unfair.
[§3.2, §3.3, Figure 3] The confidence estimator is trained and evaluated only on the MMLU dataset (Figure 3), while the evaluation tasks in Table 1 are open-ended and out-of-distribution relative to the multiple-choice MMLU regressor. The preference labels in §3.3 assume that Confidence(q,a) is positively correlated with the correctness of the initial answer a, but no calibration curve, AUROC, or accuracy-vs-confidence breakdown is reported for GSM8K, PrOntoQA, StrategyQA, CSQA2, CREAK, or BIG-Bench. The paper therefore does not establish that the high-confidence-is-correct assumption holds in the deployment setting, leaving open the possibility that confidently-wrong answers are reinforced by DPO.
[Table 1, Appendix D] All results are reported as single point estimates without error bars, standard deviations, or significance tests. Several of the reported improvements of AFICE over the strongest baseline are small, such as Vicuna on GSM8K (0.623 vs. 0.616 for Predictive Entropy) and LLaMA-3 on FOL (0.752 vs. 0.733 for Semantic Entropy). Because the underlying evaluation sets are relatively small (e.g., 464 GSM8K examples, 307 PrOntoQA examples), these differences may not be statistically reliable, and the abstract's claim of 'significant improvements' is not currently supported by appropriate statistical evidence.
[Limitations C.1, §3.3] The DPO preference labels are derived from the base model's confidence, but after fine-tuning the deployed policy has different internal states; the paper itself acknowledges in C.1 that the regressor may need to be retrained post-fine-tuning. The paper does not verify whether the fine-tuned model's behavior is actually consistent with its own post-fine-tuning confidence, which is the mechanism that the title and framing claim. This gap leaves the causal chain from confidence estimation to improved final accuracy only partially demonstrated.
[§2.1, §4.2] There is a tension between the paper's definition of 'faithful integrity' — responding according to what the model believes to be true — and the evaluation metric, which is accuracy against ground-truth labels. If the base model is confidently wrong about an answer, persisting in that answer is faithful by the paper's definition but counts as an error in the LLM False scenario. The paper does not analyze the rate of confidently-wrong initial answers on the evaluation tasks, and thus does not resolve whether the observed LLM False improvements reflect genuine correction of low-confidence errors or are partly artifacts of the synthetic user arguments and dataset composition.

minor comments (5)

[Appendix B.1] There are typos in the prompt templates, such as 'Gengerating response candidates' instead of 'Generating response candidates', which should be corrected.
[References] Mielke et al. 2022a and 2022b are the same paper and should be consolidated into a single reference to avoid confusion.
[§4.3, Figure 4] The description of the alpha and gamma sensitivity analysis is unclear: 'AUROC peaks near zero for both, declines between 0.5 and 1, and then gradually improves' needs a clearer explanation of what is being varied and what the observed trend means for the chosen default values.
[Table 2] The column labeled 'Num.' in Table 2 is not defined in the text; a note explaining that it refers to the number of output sequences generated per input would improve readability.
[Equation (2)] Equation (2) contains a stray closing brace at the end of the display, and the notation for the hyperparameters θ_h is not clearly separated from the model parameters θ in the conditional probability; this should be cleaned up.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the held-out accuracy evaluation is not a restatement of the confidence-based preference labels.

full rationale

The paper's central claim is that AFICE improves the final-answer accuracy of LLMs in argumentative conversations. The preference labels in Section 3.3 are constructed from Confidence(q,a), the model's self-estimated confidence, but the evaluation metric is final response accuracy against ground truth on held-out benchmarks (GSM8K, PrOntoQA, StrategyQA, CSQA2, CREAK, BIG-Bench). No equation identifies final accuracy with Confidence(q,a), and the BCE regressor is trained on a 20% MMLU subset to predict semantic entropy, then evaluated on the remaining 80% and on other datasets. This is an empirical generalization, not a definitional equivalence. The paper's own Appendix C.1 warns that post-finetuning internal states may require regressor retraining, which is a validity risk rather than a circular step. The self-citations to Deng et al. (2024) and Chen et al. (2023) provide background definitions and framing, but they are not load-bearing for the method or the reported results. The concern that confidently-wrong answers could be reinforced is a correctness risk, not a circularity, because the evaluation does not reduce to the confidence estimates by construction. Therefore no circular step is exhibited.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The framework introduces no new physical or conceptual entities; it relies on the model's internal states, semantic entropy, and DPO. Its contributions are the specific estimator equations and the preference construction, which depend on several fitted hyperparameters and domain assumptions.

free parameters (7)

α = 0.7
Hyperparameter in Eq. 4 controlling the mapping from semantic entropy to question-level confidence; tuned by AUROC on MMLU (Section 4.3).
γ = 0.3
Hyperparameter in Eq. 8 controlling the weight of the answer-based cumulative probability ratio; tuned by AUROC on MMLU (Section 4.3).
K = 60
Early truncation length for sampled responses; selected to balance AUROC and compute time (Section 4.3).
N = 20
Number of multinomial beam samples; selected to balance AUROC and compute time (Section 4.3).
layer index for features = 26
Layer from which hidden states are taken as regression features; no justification given other than prior practice (Section 3.2).
sampling top_p and temperature = 0.6 and 0.9
Sampling hyperparameters for generation; fixed without sensitivity analysis (Section 4.1).
confidence thresholds = 66.7th and 33.3rd percentiles of Confidence(q,a) on MMLU
Thresholds map confidence to positive/negative preference sets; computed from training data, no theoretical grounding (Section 3.3).

assumptions (5)

domain assumption Semantic entropy is a valid reference-free measure of a model's uncertainty in a response.
Used as the regression target in Stage 2 and as a baseline in evaluation (Sections 3.2 and 4.1).
domain assumption The hidden states at layer 26 of the last token contain information sufficient to predict semantic entropy.
Relies on prior findings about internal-state truthfulness (Section 3.2).
standard math DPO can effectively instill the confidence-based stance policy beyond the base model.
DPO is a known algorithm; the paper assumes the optimization transfers to the target behavior (Section 3.3).
ad hoc to paper The two-round conversation format and model-generated user arguments are a faithful operationalization of 'faithful integrity'.
The evaluation protocol is constructed by the authors and not validated against human judgments (Section 4.1).
domain assumption Confidence thresholds computed on MMLU generalize to the evaluation datasets.
The preference data uses MMLU-specific percentiles; generalization is assumed (Section 3.3).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Aligning Large Language Models for Faithful Integrity Against Opposing Argument." pith.science (2026). https://pith.science/paper/UVXB45G5

@misc{pith2026250101336,
  author       = {Pith},
  title        = {Pith review of: Aligning Large Language Models for Faithful Integrity Against Opposing Argument},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UVXB45G5}},
  note         = {Machine review of arXiv:2501.01336}
}

read the original abstract

Large Language Models (LLMs) have demonstrated impressive capabilities in complex reasoning tasks. However, they can be easily misled by unfaithful arguments during conversations, even when their original statements are correct. To this end, we investigate the problem of maintaining faithful integrity in LLMs. This involves ensuring that LLMs adhere to their faithful statements in the face of opposing arguments and are able to correct their incorrect statements when presented with faithful arguments. In this work, we propose a novel framework, named Alignment for Faithful Integrity with Confidence Estimation (AFICE), which aims to align the LLM responses with faithful integrity. Specifically, AFICE first designs a Bilateral Confidence Estimation (BCE) approach for estimating the uncertainty of each response generated by the LLM given a specific context, which simultaneously estimate the model's confidence to the question based on the internal states during decoding as well as to the answer based on cumulative probability ratios. With the BCE, we construct a conversational preference dataset composed of context, original statement, and argument, which is adopted for aligning the LLM for faithful integrity using Direct Preference Optimization (DPO). Extensive experimental results on a wide range of benchmarks demonstrate significant improvements in the LLM's ability to maintain faithful responses when encountering opposing arguments, ensuring both the practical utility and trustworthiness of LLMs in complex interactive settings. Code and data will be released via https://github.com/zhaoy777/AFICE.git

Figures

Figures reproduced from arXiv: 2501.01336 by the authors.

**Figure 1.** Two challenges of faithful integrity on LLM-generated responses: 1) faithful response against opposing argument, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** Overview of the AFICE framework. also align with truthful, logically consistent reasoning, especially when confronted with opposing arguments or fallacies. This problem is increasingly significant as LLMs are often prone to generating responses based on surface-level patterns rather than a deep understanding of content validity and truthfulness. The interaction scenarios can be classified into two main categories,… view at source ↗

**Figure 4.** Exploration of Optimal Values for Four Hyperpa [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 15 canonical work pages

[1]

Azaria, A.; and Mitchell, T. 2023. The internal state of an llm knows when its lying. arXiv preprint arXiv:2304.13734

arXiv 2023
[2]

E.; Elhage, N.; Hatfield - Dodds, Z.; Hernandez, D.; Hume, T.; Johnston, S.; Kravec, S.; Lovitt, L.; Nanda, N.; Olsson, C.; Amodei, D.; Brown, T

Bai, Y.; Jones, A.; Ndousse, K.; Askell, A.; Chen, A.; DasSarma, N.; Drain, D.; Fort, S.; Ganguli, D.; Henighan, T.; Joseph, N.; Kadavath, S.; Kernion, J.; Conerly, T.; Showk, S. E.; Elhage, N.; Hatfield - Dodds, Z.; Hernandez, D.; Hume, T.; Johnston, S.; Kravec, S.; Lovitt, L.; Nanda, N.; Olsson, C.; Amodei, D.; Brown, T. B.; Clark, J.; McCandlish, S.; O...

arXiv 2022
[3]

Chen, J.; and Mueller, J. 2023. Quantifying Uncertainty in Answers from any Language Model via Intrinsic and Extrinsic Confidence Assessment. ArXiv preprint, abs/2308.16175

arXiv 2023
[4]

Chen, L.; Deng, Y.; Bian, Y.; Qin, Z.; Wu, B.; Chua, T.; and Wong, K. 2023. Beyond Factuality: A Comprehensive Evaluation of Large Language Models as Knowledge Generators. In EMNLP 2023, 6325--6341

work page 2023
[5]

Chen, S.; Xiong, M.; Liu, J.; Wu, Z.; Xiao, T.; Gao, S.; and He, J. 2024. In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation. CoRR, abs/2403.01548

arXiv 2024
[6]

E.; Stoica, I.; and Xing, E

Chiang, W.-L.; Li, Z.; Lin, Z.; Sheng, Y.; Wu, Z.; Zhang, H.; Zheng, L.; Zhuang, S.; Zhuang, Y.; Gonzalez, J. E.; Stoica, I.; and Xing, E. P. 2023. Vicuna: An Open-Source Chatbot Impressing GPT-4 with 90\

2023
[7]

W.; Hou, L.; Longpre, S.; Zoph, B.; Tay, Y.; Fedus, W.; Li, E.; Wang, X.; Dehghani, M.; Brahma, S.; Webson, A.; Gu, S

Chung, H. W.; Hou, L.; Longpre, S.; Zoph, B.; Tay, Y.; Fedus, W.; Li, E.; Wang, X.; Dehghani, M.; Brahma, S.; Webson, A.; Gu, S. S.; Dai, Z.; Suzgun, M.; Chen, X.; Chowdhery, A.; Narang, S.; Mishra, G.; Yu, A.; Zhao, V. Y.; Huang, Y.; Dai, A. M.; Yu, H.; Petrov, S.; Chi, E. H.; Dean, J.; Devlin, J.; Roberts, A.; Zhou, D.; Le, Q. V.; and Wei, J. 2022. Scal...

arXiv 2022
[8]

Cobbe, K.; Kosaraju, V.; Bavarian, M.; Chen, M.; Jun, H.; Kaiser, L.; Plappert, M.; Tworek, J.; Hilton, J.; Nakano, R.; et al. 2021. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168

arXiv 2021

Show all 39 references

[9]

I don't know

Deng, Y.; Zhao, Y.; Li, M.; Ng, S.; and Chua, T. 2024. Don't Just Say "I don't know"! Self-aligning Large Language Models for Responding to Unknown Questions with Explanations. In EMNLP 2024, 13652--13673

2024
[10]

Evans, O.; Cotton - Barratt, O.; Finnveden, L.; Bales, A.; Balwit, A.; Wills, P.; Righetti, L.; and Saunders, W. 2021. Truthful AI: Developing and governing AI that does not lie. CoRR, abs/2110.06674

2021 arXiv
[11]

Geng, J.; Cai, F.; Wang, Y.; Koeppl, H.; Nakov, P.; and Gurevych, I. 2023. A Survey of Language Model Confidence Estimation and Calibration

2023
[12]

Geva, M.; Khashabi, D.; Segal, E.; Khot, T.; Roth, D.; and Berant, J. 2021. Did Aristotle Use a Laptop?A Question Answering Benchmark with Implicit Reasoning Strategies. Transactions of the Association for Computational Linguistics, 9: 346–361

2021
[13]

Hendrycks, D.; Burns, C.; Basart, S.; Zou, A.; Mazeika, M.; Song, D.; and Steinhardt, J. 2020. Measuring Massive Multitask Language Understanding . CoRR, abs/2009.03300

2020 arXiv
[14]

J.; Shen, Y.; Wallis, P.; Allen - Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W

Hu, E. J.; Shen, Y.; Wallis, P.; Allen - Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2022. LoRA: Low-Rank Adaptation of Large Language Models. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022 . OpenReview.net

2022
[16]

Kadavath, S.; Conerly, T.; Askell, A.; Henighan, T.; Drain, D.; Perez, E.; Schiefer, N.; Hatfield-Dodds, Z.; DasSarma, N.; Tran-Johnson, E.; Johnston, S.; El-Showk, S.; Jones, A.; Elhage, N.; Hume, T.; Chen, A.; Bai, Y.; Bowman, S.; Fort, S.; Ganguli, D.; Hernandez, D.; Jacobs...

2022 arXiv
[17]

Kuhn, L.; Gal, Y.; and Farquhar, S. 2023. Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation. In The Eleventh International Conference on Learning Representations

2023
[18]

Li, K.; Patel, O.; Vi \'e gas, F.; Pfister, H.; and Wattenberg, M. 2024 a . Inference-time intervention: Eliciting truthful answers from a language model. Advances in Neural Information Processing Systems, 36

2024
[19]

Li, S.; Deng, Y.; Cai, D.; Lu, H.; Chen, L.; and Lam, W. 2024 b . Consecutive Batch Model Editing with HooK Layers. In EMNLP 2024 , 13817--13833

2024
[20]

Lin, Z.; Trivedi, S.; and Sun, J. 2023. Generating with Confidence: Uncertainty Quantification for Black-box Large Language Models. CoRR, abs/2305.19187

2023 arXiv
[21]

Malinin, A.; and Gales, M. 2020. Uncertainty Estimation in Autoregressive Structured Prediction. In International Conference on Learning Representations

2020
[22]

Meta. 2024. The Llama 3 Herd of Models. CoRR, abs/2407.21783

2024 arXiv
[23]

J.; Szlam, A.; Dinan, E.; and Boureau, Y.-L

Mielke, S. J.; Szlam, A.; Dinan, E.; and Boureau, Y.-L. 2022 a . Reducing Conversational Agents ' Overconfidence Through Linguistic Calibration. Transactions of the Association for Computational Linguistics, 10: 857--872

2022
[24]

J.; Szlam, A.; Dinan, E.; and Boureau, Y.-L

Mielke, S. J.; Szlam, A.; Dinan, E.; and Boureau, Y.-L. 2022 b . Reducing Conversational Agents’ Overconfidence Through Linguistic Calibration. Transactions of the Association for Computational Linguistics, 10: 857–872

2022
[25]

Onoe, Y.; Zhang, M. J. Q.; Choi, E.; and Durrett, G. 2021. CREAK: A Dataset for Commonsense Reasoning over Entity Knowledge. In Thirty-fifth Conference on Neural Information Processing Systems, Datasets and Benchmarks Track

2021
[26]

Saparov, A.; and He, H. 2023. Language Models Are Greedy Reasoners: A Systematic Formal Analysis of Chain-of-Thought. In The Eleventh International Conference on Learning Representations

2023
[27]

W.; Chowdhery, A.; Le, Q

Suzgun, M.; Scales, N.; Sch \"a rli, N.; Gehrmann, S.; Tay, Y.; Chung, H. W.; Chowdhery, A.; Le, Q. V.; Chi, E. H.; Zhou, D.; et al. 2022. Challenging BIG-Bench tasks and whether chain-of-thought can solve them. arXiv preprint arXiv:2210.09261

2022 arXiv
[28]

Talmor, A.; Yoran, O.; Le Bras, R.; Bhagavatula, C.; Goldberg, Y.; Choi, Y.; and Berant, J. 2021. CommonsenseQA 2.0: Exposing the Limits of AI through Gamification. In Thirty-fifth Conference on Neural Information Processing Systems, Datasets and Benchmarks Track

2021
[29]

D.; and Finn, C

Tian, K.; Mitchell, E.; Yao, H.; Manning, C. D.; and Finn, C. 2023. Fine-tuning Language Models for Factuality. arXiv:2311.08401

2023 arXiv
[30]

Wang, B.; Yue, X.; and Sun, H. 2023. Can ChatGPT Defend its Belief in Truth? Evaluating LLM Reasoning via Debate. In Findings of the Association for Computational Linguistics: EMNLP 2023 , 11865--11881. Association for Computational Linguistics

2023
[31]

H.; Hashimoto, T.; Vinyals, O.; Liang, P.; Dean, J.; and Fedus, W

Wei, J.; Tay, Y.; Bommasani, R.; Raffel, C.; Zoph, B.; Borgeaud, S.; Yogatama, D.; Bosma, M.; Zhou, D.; Metzler, D.; Chi, E. H.; Hashimoto, T.; Vinyals, O.; Liang, P.; Dean, J.; and Fedus, W. 2022. Emergent Abilities of Large Language Models. arXiv:2206.07682

2022 arXiv
[32]

Wen, B.; Yao, J.; Feng, S.; Xu, C.; Tsvetkov, Y.; Howe, B.; and Wang, L. L. 2024. Know Your Limits: A Survey of Abstention in Large Language Models. arXiv:2407.18418

2024 arXiv
[33]

S.; Yang, S.; Zhang, T.; Shi, W.; Zhang, T.; Fang, Z.; Xu, W.; and Qiu, H

Xu, R.; Lin, B. S.; Yang, S.; Zhang, T.; Shi, W.; Zhang, T.; Fang, Z.; Xu, W.; and Qiu, H. 2024. The Earth is Flat because...: Investigating LLMs' Belief towards Misinformation via Persuasive Conversation. arXiv:2312.09085

2024 arXiv
[34]

Yang, Y.; Chern, E.; Qiu, X.; Neubig, G.; and Liu, P. 2023. Alignment for honesty. arXiv preprint arXiv:2312.07000

2023 arXiv
[35]

R.; and Cao, Y

Yao, S.; Zhao, J.; Yu, D.; Du, N.; Shafran, I.; Narasimhan, K. R.; and Cao, Y. 2023. ReAct: Synergizing Reasoning and Acting in Language Models. In ICLR 2023

2023
[36]

Yin, F.; Srinivasa, J.; and Chang, K.-W. 2024. Characterizing Truthfulness in Large Language Model Generations with Local Intrinsic Dimension. arXiv preprint arXiv:2402.18048

2024 arXiv
[37]

Zhang, X.; Peng, B.; Tian, Y.; Zhou, J.; Jin, L.; Song, L.; Mi, H.; and Meng, H. 2024. Self-Alignment for Factuality: Mitigating Hallucinations in LLMs via Self-Evaluation. arXiv:2402.09267

2024 arXiv
[38]

Zheng, D.; Liu, D.; Lapata, M.; and Pan, J. Z. 2024. TrustScore: Reference-Free Evaluation of LLM Response Trustworthiness. arXiv:2402.12545

2024 arXiv
[39]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...
[40]

write newline

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

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Azaria, A.; and Mitchell, T. 2023. The internal state of an llm knows when its lying. arXiv preprint arXiv:2304.13734

arXiv 2023

[2] [2]

E.; Elhage, N.; Hatfield - Dodds, Z.; Hernandez, D.; Hume, T.; Johnston, S.; Kravec, S.; Lovitt, L.; Nanda, N.; Olsson, C.; Amodei, D.; Brown, T

Bai, Y.; Jones, A.; Ndousse, K.; Askell, A.; Chen, A.; DasSarma, N.; Drain, D.; Fort, S.; Ganguli, D.; Henighan, T.; Joseph, N.; Kadavath, S.; Kernion, J.; Conerly, T.; Showk, S. E.; Elhage, N.; Hatfield - Dodds, Z.; Hernandez, D.; Hume, T.; Johnston, S.; Kravec, S.; Lovitt, L.; Nanda, N.; Olsson, C.; Amodei, D.; Brown, T. B.; Clark, J.; McCandlish, S.; O...

arXiv 2022

[3] [3]

Chen, J.; and Mueller, J. 2023. Quantifying Uncertainty in Answers from any Language Model via Intrinsic and Extrinsic Confidence Assessment. ArXiv preprint, abs/2308.16175

arXiv 2023

[4] [4]

Chen, L.; Deng, Y.; Bian, Y.; Qin, Z.; Wu, B.; Chua, T.; and Wong, K. 2023. Beyond Factuality: A Comprehensive Evaluation of Large Language Models as Knowledge Generators. In EMNLP 2023, 6325--6341

work page 2023

[5] [5]

Chen, S.; Xiong, M.; Liu, J.; Wu, Z.; Xiao, T.; Gao, S.; and He, J. 2024. In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation. CoRR, abs/2403.01548

arXiv 2024

[6] [6]

E.; Stoica, I.; and Xing, E

Chiang, W.-L.; Li, Z.; Lin, Z.; Sheng, Y.; Wu, Z.; Zhang, H.; Zheng, L.; Zhuang, S.; Zhuang, Y.; Gonzalez, J. E.; Stoica, I.; and Xing, E. P. 2023. Vicuna: An Open-Source Chatbot Impressing GPT-4 with 90\

2023

[7] [7]

W.; Hou, L.; Longpre, S.; Zoph, B.; Tay, Y.; Fedus, W.; Li, E.; Wang, X.; Dehghani, M.; Brahma, S.; Webson, A.; Gu, S

Chung, H. W.; Hou, L.; Longpre, S.; Zoph, B.; Tay, Y.; Fedus, W.; Li, E.; Wang, X.; Dehghani, M.; Brahma, S.; Webson, A.; Gu, S. S.; Dai, Z.; Suzgun, M.; Chen, X.; Chowdhery, A.; Narang, S.; Mishra, G.; Yu, A.; Zhao, V. Y.; Huang, Y.; Dai, A. M.; Yu, H.; Petrov, S.; Chi, E. H.; Dean, J.; Devlin, J.; Roberts, A.; Zhou, D.; Le, Q. V.; and Wei, J. 2022. Scal...

arXiv 2022

[8] [8]

Cobbe, K.; Kosaraju, V.; Bavarian, M.; Chen, M.; Jun, H.; Kaiser, L.; Plappert, M.; Tworek, J.; Hilton, J.; Nakano, R.; et al. 2021. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168

arXiv 2021