BiCon-Gate: Consistency-Gated De-colloquialisation for Dialogue Fact-Checking

Hyunkyung Park; Arkaitz Zubiaga

cs.CLApr 15, 2026

BiCon-Gate: Consistency-Gated De-colloquialisation for Dialogue Fact-Checking

Authors: Hyunkyung Park, Arkaitz Zubiaga

Organizations: Queen Mary University of London

Abstract

Automated fact-checking in dialogue involves multi-turn conversations where colloquial language is frequent yet understudied. To address this gap, we propose a conservative rewrite candidate for each response claim via staged de-colloquialisation, combining lightweight surface normalisation with scoped in-claim coreference resolution. We then introduce BiCon-Gate, a semantics-aware consistency gate that selects the rewrite candidate only when it is semantically supported by the dialogue context, otherwise falling back to the original claim. This gated selection stabilises downstream fact-checking and yields gains in both evidence retrieval and fact verification. On the DialFact benchmark, our approach improves retrieval and verification, with particularly strong gains on SUPPORTS, and outperforms competitive baselines, including a decoder-based one-shot LLM rewrite that attempts to perform all de-colloquialisation steps in a single pass.

Figures & tables

Figure 1: Example multi-turn dialogue illustrating a context-dependent response claim about the Apollo 11 mission . In-claim pronouns in the response refer to antecedents introduced in earlier turns.

Figure 2: Overview of the staged de-colloquialisation pipeline (

R_{1}

R_{5}

), BiCon-Gate routing, and evaluation protocols (IR-only, FV-only, and E2E). BiCon-Gate combines bidirectional NLI entailment/contradiction signals ( ent / cnt ) with embedding cosine similarity ( sim ) and accepts a rewrite only when the gate score exceeds a task-specific threshold (

\tau

); otherwise it falls back to

R_{0}

Figure 3: IR gate sweep on the validation split. We report BM25 micro-Recall@180 and $1{-}\mathrm{ZHR}@180$ as a function of the $R_{4}$ gate threshold $\tau$ (where $\mathrm{ZHR}@180$ is the zero-hit rate). We use $\tau_{\text{IR}}=0.50$ in the IR-only protocol.

Figure 4: FV-only gate sweep on the validation split (gold evidence). Macro-F1 and macro-Recall as a function of the $R_{4}$ gate threshold $\tau$ ; we use $\tau_{\text{FV}}=0.70$ in FV-only and E2E.

	BM25 (K=180)		E5 dense (K=10)		BGE-CE (K=1)
Claim	R@180	nDCG@180	R@10	nDCG@10	R@1	nDCG@1
$R_{0}$	73.89	48.37	70.47	61.98	53.51	57.01
$R_{1}$	73.22	47.96	69.95	61.56	53.14	56.59
$R_{2}$	73.22	47.96	69.94	61.60	53.18	56.64
$R_{3}$	73.22	47.96	69.96	61.67	53.21	56.66
$R_{4}$	76.54	50.94	73.18	64.45	55.53	59.15

Table 1: IR-only retrieval results on the test (factual) split (macro Recall/nDCG, %) for claim variants

R_{0}

–

R_{4}

R_{4}

+ Gated applies BiCon-Gate with

\tau_{\text{IR}}=0.50

and falls back to

R_{0}

when the rewrite is rejected.

Claim	Acc	Macro-F1	F1(S)	F1(R)	F1(NEI)	$\Delta$ F1
$R_{0}$	62.67	61.09	45.08	76.77	61.42	–
$R_{1}$	62.66	61.08	45.00	76.78	61.46	-0.01
$R_{2}$	62.71	61.16	45.33	76.69	61.47	+0.07
$R_{3}$	62.80	61.22	45.25	76.75	61.67	+0.13
$R_{4}$	63.15	61.85	47.59	76.20	61.77	+0.76
+ Gated	63.77	62.93	50.59	76.12	62.10	+1.84

Table 2: FV-only results on test split with gold evidence, comparing claim surfaces

R_{0}

R_{5}

(Accuracy and macro/class-wise F1, %).

R_{4}

+ Gated and

R_{5}

+ Gated apply BiCon-Gate with

\tau_{\text{FV}}=0.70

(fallback to

R_{0}

), and

\Delta

F1 is the macro-F1 change over

R_{0}

Claim	Acc	Macro-F1	F1(S)	F1(R)	F1(NEI)	$\Delta$ F1
$R_{0}$	34.85	22.60	2.45	15.99	49.37	0.00
$R_{1}$	34.86	22.62	2.50	16.01	49.37	+0.02
$R_{2}$	34.96	22.71	2.55	16.13	49.45	+0.11
$R_{3}$	35.02	22.88	2.89	16.31	49.44	+0.28
$R_{4}$	34.10	20.73	3.03	9.57	49.60	$-1.87$
+ Gated	35.54	23.22	4.56	14.84	50.24	+0.62

Table 3: End-to-end results on the full test split using the top-1 retrieved passage as evidence (Accuracy and macro/class-wise F1, %).

R_{4}

+ Gated uses BiCon-Gate with

\tau_{FV}{=}0.70

Figure 5: Effect of gating on the decoder one-shot rewrite ( $R_{5}$ ) in FV-only on the validation split. As $\tau$ increases (rejecting more $R_{5}$ instances), macro-F1 and macro-Recall approach the $R_{0}$ baseline, indicating that the gate mitigates noisy rewrites primarily via fallback to $R_{0}$ .

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Split	Type	#	S	R	NEI	Ev.Item	Turns( $C$ )	hasPronouns
Valid	factual	8,691	3,342	3,363	1,986	1.13	4.58	3,395
	personal	1,745	0	0	1,745	1.01	4.36	747
	total	10,436	3,342	3,363	3,731	1.11	4.54	4,142
Test	factual	10,420	3,939	3,935	2,546	1.32	4.35	4,549
	personal	1,389	0	0	1,389	1.10	3.80	679
	total	11,809	3,939	3,935	3,935	1.29	4.28	5,228

Appendix

Table A1: DialFact validation/test statistics and label counts. IR metrics are computed on the factual subset; FV/E2E use the full label distribution unless stated otherwise. hasPronouns counts samples containing at least one in-claim pronoun .

Protocol	Metrics
IR (gate tuning)	micro-Recall@K, $1-\mathrm{ZHR}@K$ (BM25@180)
IR (doc-level)	macro-Recall@K, nDCG@K
	(BM25@180, E5@10, BGE-CE@1)
FV-only	Accuracy, macro-F1, classwise F1 (S/R/NEI)
End-to-End	Same FV metrics as FV-only, using IR-produced evidence

Appendix

Table A2: Evaluation metrics: IR (gate tuning) metrics are computed on the valid split to select

\tau_{\mathrm{IR}}

, whereas IR (doc-level), FV-only, and E2E metrics are reported on the test split.

\mathrm{ZHR}

is the zero-hit rate; the proportion of queries for which no gold passage is retrieved in the top-K list.

Example 1 (ID: 877___8--1)

$C$	Elvis is great, he was born in 1935. … yeah he really brought rock and roll to the masses, thanks elvis. THANKS A LOT DUDE.
$R_{0}$	Remember when heartbreak hotel came out? The public hated it at first!
$R_{4}$	Remember when heartbreak hotel came out? The public hated heartbreak hotel at first! (gate $(R_{4})=0.8178$ )
$R_{5}$	Remember when “Heartbreak Hotel” came out? The public hated it at first! (gate $(R_{5})=0.5445$ )
Example 2 (ID: 445___6--2)

Appendix

Table A3: Representative DialFact instances used for qualitative analysis. Each block shows an excerpt of the dialogue context (

C

) and the resulting claim surface under the original claim (

R_{0}

), the scoped antecedent-substitution rewrite (

R_{4}

), and the decoder one-shot rewrite (

R_{5}

). Edited spans are marked with … ; BiCon-Gate scores for

R_{4}

and

R_{5}

are shown in parentheses.

Component	Model / identifier
Punctuation restoration ( $R_{2}$ )	oliverguhr/fullstop-punctuation-multilang-large Vandeghinste and Guhr (2024)
True-casing ( $R_{3}$ )	bert-base-cased Devlin et al. (2019)
Coreference resolver ( $R_{4}$ )	Maverick coreference resolver Martinelli et al. (2024)
Antecedent selector ( $R_{4}$ )	meta-llama/Llama-3.1-8B-Instruct Grattafiori et al. (2024)
Decoder rewrite ( $R_{5}$ )	Qwen/Qwen2.5-14B-Instruct Qwen et al. (2025)
NLI verifier (FV-only) / BiCon-Gate (NLI scorer)	MoritzLaurer/DeBERTa-v3-large-mnli-fever-anli-ling-wanli Laurer et al. (2024)

Appendix

Table A4: Model identifiers for third-party components used in our de-colloquialisation pipeline, retrieval, and verification experiments.

Prompt for decoder-based rewrite $R_{5}$
System
Follow the instructions exactly. Do not add or change facts.
User
You are an expert editor who rewrites informal, chatty utterances into well-formed declarative English without changing their meaning.
You will receive: (i) Context , a list of previous dialogue turns; and (ii) Response , the claim text to be normalised.
Task: Rewrite Response into New_Response by applying only the following operations.

Appendix

Table A5: Prompt used with Qwen2.5-14B-Instruct to generate the decoder-based one-shot rewrite (

R_{5}

). The prompt constrains the model to apply only surface normalisation and unambiguous pronoun substitution based on the provided context, without changing claim’s meaning.

Figure A1: FV-only heatmap of $\Delta$ macro-F1 across context turns $k\in\{0,2,4,6,8\}$ , computed relative to $R_{0}$ at the same $k$ (Table A6 ). Boxed rows highlight $R_{4}$ and $R_{4}$ + Gated , and the main setting is $k{=}2$ .

Figure A2: Class-wise FV-only $\Delta$ F1 over context turns $k\in\{0,2,4,6,8\}$ , computed from Table A6 . Colors are clipped for readability; annotated values show the true deltas, and the dotted box marks the main setting ( $k{=}2$ ).

Automated Fact-Checking

Automated fact-checking detects checkable claims, retrieves evidence, and assesses whether that evidence supports or refutes them.

See all 139 papers

Sep 3, 2026

TRILOGUE: A Trilingual Spoken Dialogue Fact-Checking Benchmark with Evidence and Paired Audio

Modern misinformation is often heard before it is read, yet fact-checking systems are still evaluated mainly on clean written claims. Spoken dialogue remains different even when systems operate on transcripts: claims may be distributed across speakers and turns, depend on prior context, and become harder to verify when Automatic Speech Recognition (ASR) errors distort the available text. Prior spoken dialogue fact-checking resources are small, English-centric, or focused on annotation rather than end-to-end benchmarking, leaving no large multilingual benchmark with paired speech and turn-level labels. We introduce TRILOGUE (TRIlingual spoken diaLOGUE fact-checking), a large-scale trilingual benchmark of source-grounded spoken dialogues in English, Russian, and Kazakh. It contains nearly 12K dialogues, 187K turns, and 390 hours of paired audio with ASR transcripts and word-level timestamp alignments across all three languages, including nearly 5K human-recorded Russian and Kazakh dialogue files. TRILOGUE supports claim check-worthiness detection, source-article evidence retrieval, and claim verification with claim-only, gold-evidence, and retrieved-evidence inputs. Baselines show that ASR degradation and cross-lingual transfer remain challenging, especially for Kazakh, while retrieved source evidence substantially narrows the gap to gold-evidence verification.

Chaewan Chun, Meruyert Aristombayeva, Jiyoung Choi +3

The Pennsylvania State University, USA · Satbayev University, Kazakhstan · Independent Researcher +1

Automated Fact-Checking Evidence Retrieval

Ask PDF

Feb 17, 2026

Fine-Refine: Iterative Fine-grained Refinement for Mitigating Dialogue Hallucination

The tendency for hallucination in current large language models (LLMs) negatively impacts dialogue systems. Such hallucinations produce factually incorrect responses that may mislead users and undermine system trust. Existing refinement methods for dialogue systems typically operate at the response level, overlooking the fact that a single response may contain multiple verifiable or unverifiable facts. To address this gap, we propose Fine-Refine, a fine-grained refinement framework that decomposes responses into atomic units, verifies each unit using external knowledge, assesses fluency via perplexity, and iteratively corrects granular errors. We evaluate factuality across the HybriDialogue and OpendialKG datasets in terms of factual accuracy (fact score) and coverage (Not Enough Information Proportion), and experiments show that Fine-Refine substantially improves factuality, achieving up to a 7.63-point gain in dialogue fact score, with a small trade-off in dialogue quality.

Xiangyan Chen, Yujian Gan, Matthew Purver

Queen Mary University of London, UK · Queen’s University Belfast, UK · Institut Jožef Stefan, Slovenia

LLM Hallucination Mitigation Hallucination Mitigation LLM Self-Refinement

Ask PDF

Jun 9, 2026

Context-Aware Multimodal Claim Verification in Spoken Dialogues

Spoken factual claims often occur within multi-turn conversations, where surrounding dialogue can provide context unavailable from the claim alone. Yet most fact-checking research evaluates isolated text, leaving conversational audio under-studied. We introduce MAD2, a synthetic Multi-turn Audio Dialogues benchmark for spoken claim verification with 1,000 two-speaker dialogues, 1,230 sentence-level check-worthy candidate annotations, and approximately 10 hours of audio. We also propose calibrated multimodal fusion of a context-aware audio encoder and a dialogue-aware text model. Adding dialogue context improves verification across settings, although the gains vary by scenario. Past-only context often approaches local offline performance, suggesting its usefulness when future context is unavailable. Fusion achieves its highest mean advantage over text with full-dialogue context, but does not consistently or significantly outperform text across settings. Under full-dialogue context, exploratory subgroup results show greater performance variation across dialogue scenarios for text and fusion, but across spread styles for audio.

Chaewan Chun, Delvin Ce Zhang, Dongwon Lee

The Pennsylvania State University, USA · University of Sheffield, UK

Claim Verification Audio Understanding Automated Fact-Checking Multi-Turn Dialogue Evaluation Multimodal Fusion Multimodal Understanding

Ask PDF

Abstract

Automated Fact-Checking

Explore similar work

TRILOGUE: A Trilingual Spoken Dialogue Fact-Checking Benchmark with Evidence and Paired Audio

Fine-Refine: Iterative Fine-grained Refinement for Mitigating Dialogue Hallucination

Context-Aware Multimodal Claim Verification in Spoken Dialogues