VARI: Robust Text-to-Video Generation via Group-Structured Intent Alignment

Mingxuan Zhu; Hohei Chan; Shiyuan Liu; Yuanxing Zhang; Lu Chen

cs.CVOct 13, 2026

VARI: Robust Text-to-Video Generation via Group-Structured Intent Alignment

Authors: Mingxuan Zhu, Hohei Chan, Shiyuan Liu, Yuanxing Zhang, Lu Chen

Organizations: Shanghai Jiao Tong University · KlingAI · South China University of Technology

Abstract

Modern text-to-video (T2V) models generate visually compelling videos, yet meaning-preserving changes in prompt wording can substantially alter semantic realization and overall quality. Existing prompt rewriting modifies the inference input, whereas conventional post-training constructs supervision from one prompt formulation at a time; neither explicitly optimizes reliability across equivalent forms. We introduce VARI (Verified Alignment over Representations of Intent), a group-structured framework that treats automatically verified formulations of the same visual intent as a shared supervision unit. VARI builds matched form-seed rollout pools, cross-scores each video against all forms, and applies a wording-independent structured intent audit to identify valid realizations and hard semantic failures. Group-Consistent Distillation (GCD) aligns every verified form with interchangeable valid realizations, while Form-Contrastive Flow-DPO (FC-DPO) learns group-level preferences from audited positive-negative pairs, prioritizing same-seed, different-form contrasts. On HunyuanVideo-1.5, VARI attains the highest T2V-CompBench, VBench, and StoryEval scores among the evaluated methods. On prompt-form extensions of T2V-CompBench and VBench, it improves weakest-form alignment over the strongest prompt-level baselines by 2.45 and 3.55 points and reduces cross-form drop by 30.5% and 32.2%, respectively, while preserving visual quality. Human evaluation, natural-caption controls, prompt-enhancement experiments, diversity analyses, and transfer to LTX-2.3 further support the robustness and transferability of these gains, while VARI requires no additional inference-time module.

Figures & tables

Figure 1: Overview of VARI. (A) Verified forms and shared seeds define a rollout grid. Cross-scoring provides mean support $\mu$ , weakest-view support $w$ , dispersion $d$ , and quality $Q$ ; a structured audit against $T_{z}$ assigns semantic type, and GroupScore ranks candidates within type. (B) GCD distills retained positives across all forms, while FC-DPO lifts audited intent-level preferences to the full group, prioritizing same-seed, different-form pairs.

Method	VBench Total	VBench Qual.	VBench Sem.	T2V-C Avg.	StoryEval
Base	80.11	82.83	69.23	47.69	32.11
VideoDPO	81.06	83.68	70.59	49.60	34.62
OnlineVPO	81.60	84.19	71.24	50.49	35.31
GCD	81.15	83.46	71.91	50.48	35.73
VARI	82.67	84.91	73.71	52.13	37.58

Table 1: Standard benchmark results. T2V-C averages seven official categories; StoryEval reports strict event completion.

Method	Cons. Attr.	Dyn. Attr.	Spatial Rel.	Motion Bind.	Action Bind.	Interaction	Numeracy	Avg.
Base	79.82	7.47	53.11	40.29	69.31	60.57	23.24	47.69
VideoDPO	81.34	8.25	55.56	41.89	71.24	63.49	25.42	49.60
OnlineVPO	82.31	10.04	56.72	43.91	72.13	64.49	23.81	50.49
GCD	83.26	8.66	56.97	39.88	72.71	66.61	25.29	50.48
VARI	86.64	9.58	58.55	40.42	75.06	67.78	26.87	52.13

Table 2: T2V-CompBench categories. VARI gives the best average; OnlineVPO leads dynamic attribute and motion binding.

Method	T2V-CompBench-R				VBench-R Semantic				VBench-R Quality
Method	Mean $\uparrow$	Worst $\uparrow$	Drop $\downarrow$	FIFR $\downarrow$	Mean $\uparrow$	Worst $\uparrow$	Drop $\downarrow$	FIFR $\downarrow$	Mean $\uparrow$
Base	47.18	44.47	4.92	11.14	68.29	61.27	11.74	11.38	82.78
VideoDPO	49.54	47.08	3.97	8.93	70.16	65.18	9.19	8.72	83.51
OnlineVPO	50.09	46.95	4.12	8.21	70.79	65.07	8.61	8.26	83.94
GCD	50.39	48.81	2.52	5.50	71.36	67.12	6.51	6.08	83.47
VARI	51.25	49.53	2.76	4.43	72.68	68.73	5.84	5.27	84.24

Table 3: Prompt-form group extensions of the official evaluators; VBench quality is reported separately.

Figure 2: Representative mining-view score matrices for one intent group. Rows are generated candidates and columns are verified scoring forms. Cell variation is diagnostic evidence, not form-dependent ground truth for a fixed video.

Figure 3: Group-level VBench-R semantic distributions. VARI raises worst-form scores and reduces cross-form drops.

Figure 4: Source/equivalent-form cases. Baselines violate spatial relation or action binding; VARI preserves both across forms. The panels show temporally ordered frame strips from the evaluated video clips.

Appendix figures & tables29 assets

Supplementary material from the paper’s appendix.

Appendix

Family	Allowed surface variation
Lexical	Synonyms and local wording changes without semantic strengthening
Syntactic	Clause reordering, attachment changes, or role-preserving voice conversion
Caption-like	Natural descriptive-caption syntax with the same commitments
Concise	Removal of linguistic redundancy while retaining every visual constraint
Relation-focused	Reordering that foregrounds an existing spatial or interaction relation
Action-focused	Reordering that foregrounds an existing action or event sequence

Appendix

Table 6: Prompt-form transformation families. A candidate is rejected whenever its visual content changes, regardless of its nominal family.

Compiler stage	Count	Rate (%)
Generated candidates	53,487	100.00
Malformed / near-duplicate removed	2,713	5.07
Field-wise verifier inputs	50,774	94.93
Verifier-accepted candidates	34,318	64.16
Retained rewrites (four per source)	20,000	37.39
Complete five-form groups	5,000	100.00 intents

Appendix

Table 7: Prompt-form generation and verification funnel. Candidate-stage rates use 53,487 as the denominator; complete-group coverage is reported over 5,000 source intents.

Critical field	Equivalent / total	Precision (%)
Entity / count	59/60	98.33
Spatial relation	57/60	95.00
Action–role binding	58/60	96.67
Temporal order	56/60	93.33
Attribute / state / other	58/60	96.67
Overall	288/300	96.00

Appendix

Table 8: Human semantic-equivalence audit of accepted prompt forms.

Audit subset	Equivalent / total	Precision (%)
Training accepted pairs	190/200	95.00
Benchmark accepted pairs	98/100	98.00
All accepted pairs	288/300	96.00
Complete benchmark groups (all four rewrites valid)	97/100	97.00

Appendix

Table 9: Source-specific pair audit and complete-group benchmark audit. Group success requires all four rewrites to preserve the shared target.

Compiler outcome	Count	Rate (%)
At least one audited positive	4,371	87.42
At least one audited negative	3,228	64.56
At least one preference pair	3,069	61.38
GCD-only intent	1,302	26.04
No retained supervision	629	12.58
Matched-seed retained pairs	4,236	71.71

Appendix

Table 10: Supervision yield. Intent rates use 5,000 source intents; pair-provenance rates use 5,907 retained preference pairs.

Commitments	$N$	Positive	Negative	Pair
$J_{z}=3$ – $4$	1,617	1,492 (92.3)	913 (56.5)	876 (54.2)
$J_{z}=5$ – $6$	2,356	2,068 (87.8)	1,539 (65.3)	1,474 (62.6)
$J_{z}\geq 7$	1,027	811 (79.0)	776 (75.6)	719 (70.0)
All intents	5,000	4,371 (87.4)	3,228 (64.6)	3,069 (61.4)

Appendix

Table 11: Supervision yield by the number of atomic commitments. Cells report count (rate %). Min/max audit aggregation is deliberately more conservative for longer structured targets.

Item	Data and generation configuration
Main / transfer backbone	HunyuanVideo-1.5 / LTX-2.3 (22B dev)
Trainable component	Video DiT only; text encoders and VAE frozen
Training intents	5,000
Prompt-form groups	One source form and four verified semantic-preserving rewrites per intent
Generation seeds	Four intent-specific seeds reused across all verified forms
Initial rollout pool	100,000 videos ( $5{,}000$ intents $\times$ 5 forms $\times$ 4 seeds)

Appendix

Table 12: Core implementation configuration. Compared methods use matched generation settings and total sampling budgets whenever permitted by the method definition.

Method	Rollouts	Scoring / supervision	Optimization pattern	Inference overhead
VideoDPO	100k	Single-form reward and pair construction	Fixed-pool preference training	None
OnlineVPO	100k	Online reward and pair refresh	Interleaved generation and updates	None
VARI	100k	Cross-form scores; shortlist-only audit	Parallel offline GCD + FC-DPO	None

Appendix

Table 13: Per-run protocol comparison under a matched video-rollout budget. API calls for the shortlist audit are logged separately by the compiler.

T2V-C category	Consist.	Dynamic	Spatial	Motion	Action	Interact.	Numeracy
$\tau_{b}$ (%)	78.07	7.18	50.22	36.31	65.09	56.37	20.19

Appendix

Table 14: Validation-fixed success thresholds used by T2V-CompBench-R FIFR. VBench-R analogously uses the base-model validation median of each matched semantic dimension.

Method	T2V-C Avg.	T2V-C-R Worst	VBench-R Worst
OnlineVPO	$50.49\!\pm\!0.23$	$46.95\!\pm\!0.27$	$65.07\!\pm\!0.29$
GCD	$50.48\!\pm\!0.18$	$48.81\!\pm\!0.21$	$67.12\!\pm\!0.25$
VARI	$\textbf{52.13}\!\pm\!0.19$	$\textbf{49.53}\!\pm\!0.22$	$\textbf{68.73}\!\pm\!0.26$

Appendix

Table 15: Mean

\pm

standard deviation over three independent training runs.

Attribution-critical variant	Mean	Worst	Drop	FIFR
GCD	$50.46\!\pm\!0.18$	$48.88\!\pm\!0.21$	$2.57\!\pm\!0.13$	$5.41\!\pm\!0.24$
Intent-audit-only FC-DPO	$50.94\!\pm\!0.15$	$49.09\!\pm\!0.19$	$2.61\!\pm\!0.12$	$5.02\!\pm\!0.22$
VARI w/o matched-seed pairs	$51.06\!\pm\!0.19$	$49.28\!\pm\!0.18$	$2.46\!\pm\!0.16$	$4.87\!\pm\!0.20$
VARI (matched-seed prioritized)	$\textbf{51.17}\!\pm\!0.13$	$\textbf{49.51}\!\pm\!0.16$	$\textbf{2.29}\!\pm\!0.11$	$\textbf{4.44}\!\pm\!0.18$

Appendix

Table 16: Mean

\pm

standard deviation over three paired training branches on the fixed development split.

Mined type	Human precision	Visual usability	Indep. agreement
Group-consistent positive	96/100	91/100	88/100
Low-score negative	76/100	47/100	68/100
Dispersion-only negative	58/100	81/100	50/100
Intent-audit negative	85/100	76/100	78/100
Audited FC negative	91/100	79/100	84/100

Appendix

Table 17: Blind supervision audit. Bold marks the best result among negative-mining variants; each cell reports valid counts out of

N=100

candidates per type.

Evaluation source	Retrieved candidates	Confirmed shared intents
VBench	13	0
T2V-CompBench	23	0
StoryEval	8	0
VATEX	17	0
MSR-VTT	26	0

Appendix

Table 18: Training–evaluation overlap audit. Exact normalized overlap is zero for every source.

Method	Appearance div.	Motion div.	CF margin
Base	100.00	100.00	11.36
OnlineVPO	98.20	97.40	12.01
GCD	95.70	95.30	12.71
VARI	96.50	95.80	13.90

Appendix

Table 19: Diversity retention relative to Base (%) and counterfactual semantic margin.

Change type	Base	OnlineVPO	GCD	VARI
Count	10.77	11.41	11.68	13.12
Spatial relation	13.26	14.01	14.82	15.97
Action-role swap	11.88	12.59	13.37	14.53
Temporal order	9.52	10.04	10.96	11.98

Appendix

Table 20: Counterfactual margin by minimally changed semantic commitment.

Category	Base	VideoDPO	OnlineVPO	GCD	VARI
Consistency	79.82	81.34	82.31	83.26	86.64
Dynamic	7.47	8.25	10.04	8.66	9.58
Spatial	53.11	55.56	56.72	56.97	58.55
Motion	40.29	41.89	43.91	39.88	40.42
Action	69.31	71.24	72.13	72.71	75.06
Interaction	60.57	63.49	64.49	66.61	67.78

Appendix

Table 21: T2V-CompBench category scores under the unified evaluation protocol. Values are percentages; the table is transposed to preserve a stable single-column layout.

Variant	Mean	Worst	Drop	FIFR
Reward-SFT	48.58	43.82	5.97	13.11
Prompt-DPO	49.91	46.77	4.49	9.24
FC-DPO	50.31	47.68	4.01	7.76
GCD	50.46	48.88	2.57	5.41
GCD + prompt-DPO	50.82	48.64	3.28	6.19
VARI	51.17	49.51	2.29	4.44

Appendix

Table 22: Training-objective ablation on T2V-CompBench-R.

Variant	Mean	Worst	Drop	FIFR
Reward-SFT	48.58	43.82	5.97	13.11
Prompt-level DPO	49.91	46.77	4.49	9.24
Multi-form prompt-DPO	50.08	47.03	4.34	8.61
Direct worst-form DPO	50.39	48.17	3.12	6.38
Form-specific distillation	49.93	47.51	3.69	7.19
Single-positive GCD	50.18	48.56	2.89	5.93

Appendix

Table 23: Full causal ablation on the fixed T2V-CompBench-R development split.

Mining score	$\mu$	$w$	$-d$	Worst $\uparrow$	Drop $\downarrow$	FIFR $\downarrow$
Single-form	✗	✗	✗	41.08	16.89	14.94
Mean only	✓	✗	✗	44.17	13.43	11.18
Mean + worst	✓	✓	✗	46.12	10.21	8.24
Mean + worst - drop	✓	✓	✓	46.95	7.48	5.96
Full + $Q$	✓	✓	✓	46.87	7.64	6.08

Appendix

Table 24: GroupScore-component ablation with T2V-CompBench-R evaluation.

Setting	Worst	Drop	FIFR
Random positive vs random negative	43.12	15.79	14.09
Best-group vs low-score	45.07	12.34	9.88
Best-group vs worst-group	45.74	9.83	7.41
Best-group vs form-contrastive	46.88	7.54	5.89
Top- $m$ positives + mixed negatives	46.63	7.82	5.74
1 form, 4 seeds	41.23	16.21	14.98

Appendix

Table 25: Pair-construction diagnostics (upper block) and rollout-scale diagnostics (lower block).

Group construction	Mean	Worst	Drop	FIFR
Single original form	49.91	41.24	16.23	14.98
Two verified forms	50.55	43.87	11.82	10.69
Five unverified rewrites	51.06	43.39	13.14	11.91
Five verified forms	51.76	46.81	7.74	6.04

Appendix

Table 26: Effect of prompt-form verification and group size.

Input	Generator	T2V-C Avg.	R-Worst	R-Drop	R-FIFR
Raw	Base DiT	47.69	44.47	4.92	11.14
Raw	VARI DiT	52.13	49.53	2.76	4.43
Official PE	Base DiT	57.63	53.09	6.91	7.64
Official PE	VARI DiT	59.91	56.28	4.57	4.14

Appendix

Table 27: Prompt-enhancement compatibility under a shared enhanced input.

LTX-2.3	T2V-C	R-Worst	R-Drop	VBench-R	Story
Base	44.79	41.86	5.48	58.31	29.66
VideoDPO	46.31	43.19	4.41	60.08	31.02
OnlineVPO	46.84	43.04	4.82	60.67	31.48
+GCD	47.57	44.79	3.51	61.61	32.17
+VARI	48.29	45.76	3.01	63.21	33.37

Appendix

Table 28: Cross-backbone validation on LTX-2.3. VBench-R reports semantic Worst; Story denotes StoryEval.

Method	Base	Reward-SFT	VideoDPO	OnlineVPO	VARI
StoryEval strict score	32.11	33.21	34.62	35.31	37.58

Appendix

Table 29: StoryEval strict event-completion score.

Method	Mean	Worst	Drop	FIFR
Base	63.77	59.66	6.18	9.21
VideoDPO	65.99	62.02	5.31	7.43
OnlineVPO	66.47	61.92	6.14	7.75
GCD	67.29	64.49	4.17	5.33
VARI	68.16	65.09	4.03	4.85

Appendix

Table 30: Verified VATEX multi-caption diagnostic scored with held-out VisionReward.

Method	Mean	Worst	Drop	FIFR
Base	61.63	57.17	6.67	10.22
VideoDPO	63.61	59.33	6.07	8.76
OnlineVPO	64.05	59.24	6.37	9.00
GCD	65.21	61.69	4.96	6.33
VARI	65.88	62.34	4.69	5.84

Appendix

Table 31: Disjoint MSR-VTT control under the same held-out reward protocol.

Dimension	Counts			Raw shares (%)			W–L	95% CI	$\alpha$
	W	T	L	W	T	L	(pt.)
Intent faithfulness	352	96	152	58.67	16.00	25.33	33.33	[25.9, 40.6]	0.59
Cross-form consistency	378	83	139	63.00	13.83	23.17	39.83	[32.4, 47.1]	0.63
Visual quality	319	101	180	53.17	16.83	30.00	23.17	[14.9, 31.3]	0.57
Overall preference	347	106	147	57.83	17.67	24.50	33.33	[26.3, 40.4]	0.60

Appendix

Table 32: Detailed blinded human evaluation: VARI versus OnlineVPO. Each dimension contains 600 judgments.

Figure 5: Four additional held-out cases with ordered frames. (a) VARI preserves five salient robots and the group dance across both forms. (b) It maintains one coherent game layout and stable participants. (c) It moves the car past a stable church. (d) It realizes progressive snowman collapse and volume loss.

Comparison	Forms	Material shown
Main-paper spatial relation	$2\times 3$	Displayed strips (6 evaluated clips)
Main-paper action binding	$2\times 3$	Displayed strips (6 evaluated clips)
Appendix paired-form cases	$2\times 2$	Static temporally ordered panels
Appendix dynamics cases	$2\times 1$	Static temporally ordered panels

Appendix

Table 33: Qualitative evidence shown in this preprint. Each count is prompt forms

\times

compared methods or intents, as applicable.

Explore similar work

TIDE: Trajectory-Informed Skill Distillation for Efficient Text-to-Video Prompt Optimization
Oct 10, 2026Chengxing Zhou, Yuliang Huang, Jingwei SongText-to-Video Generation
Sun Yat-sen University, China · The University of Hong Kong, China
Plan-and-Verify Video Reward Reasoning with Spatio-Temporal Scene Graph Grounding
Jun 10, 2026Hyomin Kim, Junghye Kim, Joanie Hayoun Chung +4Reward Modeling Text-to-Video Generation
Department of Artificial Intelligence, Korea University · Department of Statistics, Korea University
Retrieval, Refinement, and Ranking for Text-to-Video Generation via Prompt Optimization and Test-Time Scaling
Mar 2, 2026Zillur Rahman, Alex Sheng, Cristian MeoText-to-Video Generation Video Diffusion Models
Algoverse AI

Oct 10, 2026J/K move · Enter open · S save

TIDE: Trajectory-Informed Skill Distillation for Efficient Text-to-Video Prompt Optimization

Chengxing Zhou, Yuliang Huang, Jingwei Song

Sun Yat-sen University, China · The University of Hong Kong, China

Text-to-Video Generation

Open paper AskPDF arXiv

Text-to-video (T2V) models are highly sensitive to prompt specificity, while iterative prompt optimization typically improves generation quality only through repeated video generation and vision-language model evaluation at inference time. We propose TIDE, a framework that distills this costly search process into an interpretable and reusable prompt-rewriting Skill. TIDE formulates prompt rewriting as a set of source-conditioned rule modules, each targeting a distinct semantic dimension. Once distilled, the frozen Skill activates the relevant rules and generates both positive and negative prompts in a single LLM call, followed by only one T2V generation, without test-time visual search or backbone fine-tuning. Evolved using only 100 prompts over 14 hours, TIDE achieves prompt-optimization performance comparable to computationally expensive search-based methods. Moreover, when transferred to previously unseen video-generation models, TIDE outperforms existing approaches specifically designed or trained for those models. Code is available at https://github.com/xiningin/TIDE_T2V.

Abstract

Explore similar work

TIDE: Trajectory-Informed Skill Distillation for Efficient Text-to-Video Prompt Optimization

Plan-and-Verify Video Reward Reasoning with Spatio-Temporal Scene Graph Grounding

Retrieval, Refinement, and Ranking for Text-to-Video Generation via Prompt Optimization and Test-Time Scaling