HKGAI-V1: Towards Regional Sovereign Large Language Model for Hong Kong

Junqi Zhu; Ruiyuan Zhang; Sirui Han; Yike Guo

REVIEW 4 major objections 4 minor 65 references

HKGAI-V1: Towards Regional Sovereign Large Language Model for Hong Kong

T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read HKGAI-V1 is a 685-billion-parameter model, fine-tuned from DeepSeek, that outperforms general-purpose models on Hong Kong-specific culturally sensitive queries and regional value benchmarks while keeping general knowledge and reasoning…

desk verdict A deployed Hong Kong sovereign-AI system with standard alignment methods and a self-authored benchmark suite; the regional-superiority claim is plausible but not yet established. read the letter →

arxiv 2507.11502 v1 pith:5XMU27NK submitted 2025-07-14 cs.CL cs.CEcs.LG

Sirui Han , Junqi Zhu , Ruiyuan Zhang , Yike Guo This is my paper

classification cs.CLcs.CEcs.LG

keywords sovereignAIlargelanguagemodelHongKongvaluealignmentRLHFretrieval-augmentedgenerationCantoneseregionalbenchmark

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

This paper sets out to show that a large language model can be made genuinely regional: a 685-billion-parameter model fine-tuned from DeepSeek, HKGAI-V1 is aligned to Hong Kong's Cantonese-Mandarin-English linguistic environment and its local legal, cultural, and ethical norms. The authors claim that this regional alignment improves performance on Hong Kong-specific sensitive and safety questions while leaving general knowledge and reasoning essentially intact, with notable gains on benchmarks that reward local values (88.95 vs 70.41 on Beaver-zh-hk, and 68.06 vs 30.12 on Flames) and a new state of the art on the author-built HKMMLU (81.4%). They also contribute a reusable evaluation tool, the Adversarial HK Value Benchmark, and a RAG architecture for grounded, time-sensitive answers. If the claims hold, they would show that regional digital sovereignty need not require training from scratch, and that a carefully fine-tuned open base model can outperform major general-purpose systems on locally defined criteria.

What carries the argument

The correction-amplified RLHF loop is the load-bearing mechanism. Local annotators produce a question-answer-correction dataset; the correction model HKValue-Aligner learns to rewrite unsafe or misaligned answers into compliant ones; those corrected answers become preference pairs for RLHF with a KL penalty to the base policy. Around this sits a modular RAG pipeline (intent classifier, retriever, memory, tool use, workflow moderation) and the paper's own evaluation instruments, especially the Adversarial HK Value Benchmark, which scores responses to 300 sensitive questions as safe, template refusal, or unsafe.

What would settle it

Hold out an independently authored set of Hong Kong sensitive questions, written after the model's training data was fixed and in colloquial Cantonese, and compare HKGAI-V1 with DeepSeek-R1 under identical conditions with retrieval turned off; if the Beaver-zh-hk gap of 18.54 points collapses, the regional-alignment claim is mostly benchmark-specific.

Watch

Extended reading notes

Core claim

The central discovery is that full-parameter fine-tuning of a strong general-purpose base model, guided by local annotator corrections and RLHF, can encode region-specific values without a general-capability tax. HKGAI-V1 beats DeepSeek-R1 on Beaver-zh-hk by 18.54 points and on Flames by 37.94 points, achieves 81.4% on HKMMLU versus 76.6% for DeepSeek-V3, and is rated safe on 79% of the paper's Hong Kong sensitive adversarial questions, compared with ChatGPT's 10.7%. The authors interpret these results as evidence that a sovereign model can control its own safety behavior, language matching, and local knowledge while remaining competitive on MMLU and AGI-Eval.

Load-bearing premise

The paper's headline results rest on benchmarks the authors built themselves, so the central assumption is that those benchmarks genuinely represent Hong Kong values and are not contaminated by the fine-tuning data.

Editorial extensions

If this is right

Hong Kong public-sector deployments can use a locally controlled model that answers sensitive local questions directly, without ideological templates or refusal, while staying compliant with local law.
A region can obtain a sovereign model by fine-tuning an existing open base model on local data, cutting the compute and data costs of training from scratch.
Regional value alignment does not have to trade away general knowledge: MMLU drops by only 0.36 points while regional safety scores rise sharply.
The same evaluation recipe (a local knowledge benchmark, a legal safety benchmark, and an adversarial value benchmark) can be transplanted to other jurisdictions.
The 16.5% unsafe rate on instruction attacks is a known consequence of regional alignment and defines the next target for the V2 roadmap.

Reading between the lines

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

If the benchmark results generalize beyond the paper's own test sets, then value alignment for a region can be treated as a correction-and-preference amplification problem rather than a pre-training problem, which would lower the barrier for other mid-sized jurisdictions.
The comparison on the Adversarial HK Value Bench mixes refusal with safety: ChatGPT's 10.7% 'safe' rate on Hong Kong sensitive questions could partly reflect a different refusal policy rather than a genuine lack of alignment, so the headline gap is likely smaller than it appears.
A natural next experiment is to run the same pipeline with Cantonese-only prompts and code-mixed vernacular text, since the paper's strongest language-following results are for written Chinese and English while oral Cantonese still shows search-dependent variation.
The governance-embedded architecture, which separates trust, platform, model, and data layers, could be reused as a public-sector template even if the model itself remains based on DeepSeek.

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

Colleague, here's the short version: HKGAI-V1 is a legitimate systems paper—a 685B DeepSeek fine-tune with RLHF/DPO, correction learning, and RAG, deployed to ~20,000 Hong Kong government users. That deployment is real, and the authors are candid that the model is a fine-tune, not a from-scratch sovereign model. The paper is most useful as a case study in regional alignment practice, not as a methodological advance.

What it does well: it documents a complete pipeline, including the HKValue-Aligner correction model and a set of region-specific benchmarks (HKMMLU, SafeLawBench, NaVAB, Adversarial HK Value Bench). The limitations sections (§6.1, §6.3) are honest: they acknowledge the subjectivity of 'Hong Kong values', the limited local corpus, and the model's dependence on DeepSeek. That candor counts for something.

The soft spots are in the evaluation. The central claim—superiority on Hong Kong-specific tasks—rests entirely on self-authored, unreleased benchmarks. There are no error bars, no significance tests, no contamination analysis, and no independent scoring. The 100-question sensitive-query evaluation (Table 7) is reported as aggregate percentages with no examples and no inter-annotator agreement. That's a real evidential gap. There's also an internal tension: §4.5 says the 'pre-training corpus' is extensively populated with Hong Kong text, but §6.3 says HKGAI-V1 is only a full-parameter fine-tuned version of DeepSeek. Those two statements don't reconcile, and the zero-shot framing in §4.5 is misleading given the model was fine-tuned on HK data.

The reader's rejection is a bit harsh in tone but the substance holds: the paper cannot be verified as a scientific contribution until the benchmarks are released or an external evaluation is done. That said, this isn't fraud—it's an under-supported applied claim. The paper deserves a serious referee, not because the results are convincing, but because the topic (region-specific alignment and its evaluation) is important and the deployment gives it real-world weight. A good referee could push for benchmark release, a contamination check, and a more careful separation of training and test data.

For a reading group, it's a decent case study on evaluation pitfalls. I wouldn't cite it for the specific numbers, but I might cite it as an example of deployed sovereign AI. Verdict: reject in current form, but engage with it; the revisions are feasible.

Referee Report

4 major / 4 minor

Summary. The paper reports HKGAI-V1, a 685-billion-parameter model obtained by full-parameter fine-tuning of DeepSeek, combined with RLHF, language-feedback correction via a custom HKValue-Aligner, and a modular retrieval-augmented generation (RAG) system, with the stated goal of regional value alignment for Hong Kong. The authors claim that HKGAI-V1 outperforms general-purpose models on Hong Kong-specific knowledge and safety benchmarks (HKMMLU, SafeLawBench, NaVAB, the Adversarial HK Value Benchmark, and a sensitive-question test set) while preserving general knowledge and reasoning (MMLU, AGI-Eval). The paper also discusses governance, deployment in Hong Kong government services, and a roadmap for a future HKGAI-V2.

Significance. If the empirical claims were fully substantiated, the work would be a valuable case study in region-specific alignment and sovereign LLM development, particularly for multilingual settings with Cantonese, Mandarin, and English, and for legal safety in a distinct socio-legal context. The paper is candid about several limitations in Sections 6.1 and 6.3, including annotator subjectivity, a compliance-robustness trade-off, and the model's dependence on DeepSeek. However, the central claim of superiority on Hong Kong-specific tasks rests on proprietary, unreleased benchmarks with no contamination analysis, no independent scoring, and no statistical uncertainty quantification. As presented, the contribution is primarily a system-building narrative; the measurement of regional alignment superiority is not yet established.

major comments (4)

[§4.5 vs §6.3] There is a load-bearing internal inconsistency between the HKMMLU discussion and the stated model provenance. Section 4.5 attributes the zero-shot HKMMLU result to a pre-training corpus 'extensively populated' with Hong Kong-relevant Traditional Chinese text, but Section 6.3 explicitly states that HKGAI-V1 is a full-parameter fine-tuned version of DeepSeek and that its core capabilities are predetermined by the original training. Since the authors did not pre-train the model, the zero-shot HKMMLU performance must be attributed to fine-tuning data, and the paper does not establish that the fine-tuning data and the HKMMLU test items are disjoint. This affects the central claim of regional superiority and must be resolved by clarifying data provenance and providing contamination analysis.
[§4.5, §5.3] The strongest claimed results come from self-authored, unreleased benchmarks: HKMMLU, SafeLawBench, NaVAB, and the Adversarial HK Value Benchmark. The paper provides no contamination checks, no item-level examples, no inter-annotator agreement, and no external or blinded scoring. Given that Section 4.3 describes constructing a local Hong Kong-specific Q-A dataset and generating preference data from it, the boundary between training distribution and test distribution is not demonstrated. The margins reported (e.g., HKMMLU 81.4 vs 76.6; Table 5 HK-sensitive 79% vs ChatGPT 10.7%) cannot be interpreted as genuine regional alignment until the benchmarks are released or a held-out external evaluation with contamination control is provided.
[§5.3, Table 7] The sensitive-question evaluation is based on a 100-question test set with no item examples, no scoring rubric details, no blinding procedure, and no inter-annotator reliability. Several reported metrics are non-quantitative labels ('Yes', 'Partially', 'No'), and claims such as 0% refusal and 100% positive/neutral responses are presented without confidence intervals or significance tests. Similarly, Table 5 reports percentages for the Adversarial HK Value Benchmark but does not state the number of questions per module, how the 300 total questions are allocated, or how human evaluators were trained and blinded. These omissions are load-bearing because the paper's central claim depends on these evaluations.
[§4.4, §5.3] All benchmark comparisons are single-point estimates with no error bars, repeated runs, or significance testing. The Beaver-zh-hk evaluation relies on GPT-4o scoring with a four-tier methodology that is not specified, so the robustness of the 88.95 vs 70.41 margin is unclear. Additionally, the Flames improvement (68.06 vs 30.12) is attributed to RAG and retrieval modules, but no ablation distinguishes the contribution of RAG from that of the fine-tuning and alignment pipeline; Table 6 only provides with/without-search language-following rates, not task performance. The paper should either add ablations or temper the causal claim about RAG.

minor comments (4)

[§4.1] In the RLHF objective, the maximization is written over the reward-model parameters 'phi' while the expression contains the policy parameters 'theta'; this should be 'max_theta' for consistency with the preceding sentence.
[§5.3, Table 4] The Instruction Following row says 'two languages' but then lists four (Simplified Chinese, Traditional Chinese, English, Cantonese); correct the count or split the row.
[Throughout] There are several typographical errors, including 'demonstar' (§6.1), 'standtards' (§5.3), 'Multiple-Choise' (Table 2 note), and 'endowered' (Conclusion). A copyediting pass is needed.
[References] References [31] and [66] appear to be the same arXiv paper (arXiv:2504.12911), and the acronym NaVAB is never expanded; add the full name and deduplicate the references.

Circularity Check

2 steps flagged · score 6.0 of 10

Regional-superiority claim rests on a benchmark used to refine training plus self-authored benchmarks with no contamination analysis; external benchmarks only partially mitigate.

fitted input called prediction [Section 5.3, 'Adversarial Value Bench' and Table 5]
"The HKGAI-V1 adversarial value benchmark employs 300 human-crafted sensitive questions with opposing "safe" and "unsafe" viewpoints to rigorously evaluate model alignment. ... (5) providing feedback to refine training strategies for improved safety and ethical alignment."

The benchmark is not a held-out test: its own methodology feeds model outputs and human evaluations back into training-strategy refinement, and the same benchmark's Table 5 scores are then presented as evidence that HKGAI-V1 outperforms Kimi and ChatGPT. The reported 79% safe-response rate is thus a measurement of a target the model was explicitly optimized against, so the 'prediction' of superiority is partly forced by the optimization loop rather than independently established. The paper gives no indication that a separate held-out split of the 300 questions was used for the reported numbers.
self citation load bearing [Section 4.5, 'Proprietary Evaluation Framework for HKGAI-V1' and Table 3]
"We further evaluate HKGAI-V1 on three proprietary benchmarks—HKMMLU (Hong Kong Massive Multitask Language Understanding) [7], SafelawBench [65] and NaVAB[66]. These benchmarks are specifically designed to assess a model's grasp of information relevant to Hong Kong."

The central regional-superiority claim is carried by these three benchmarks, and the cited benchmark papers ([7], [65], [66]) are authored inside the same HKGAI/HKUST group as the present paper. The paper also describes in §4.3 a 'local Hong Kong-specific Q-A dataset' used to build RLHF preference data; no contamination analysis is provided to show that the proprietary benchmark items are disjoint from that training/correction data. Consequently, the SOTA HKMMLU result (81.4%) and SafeLawBench/NaVAB margins rest on instruments whose design and scoring trace back to the same authors, making the evidence chain self-referential rather than externally falsifiable.

full rationale

The clearest circular element is the Adversarial HK Value Bench: the paper explicitly says the benchmark's outputs are used to refine training strategies, and then reports the same benchmark's scores as evidence of superiority. That is a fitted input presented as a prediction. The HKMMLU/SafeLawBench/NaVAB results are load-bearing self-citations because the benchmark papers and the model come from the same group, and the paper provides no contamination or independence analysis relative to the HK-specific Q-A and correction data used in §4.3. This is not an accusation of fraud, but it means the headline 'regional superiority' claim is at least partly an artifact of evaluation design. Some independence remains: MMLU and AGI-Eval are external and show no capability collapse, and Beaver-zh-hk and Flames are also external benchmarks where HKGAI-V1 shows large gains. These external results keep the paper from being wholly circular, so a score of 6, rather than 8 or 10, is appropriate.

Assumptions & free parameters 3 free parameters · 4 assumptions · 2 invented entities

The paper rests on standard RLHF assumptions plus the domain assumption that self-created Hong Kong benchmarks faithfully measure local values. It introduces no released artifacts, so the performance claims cannot be independently confirmed.

free parameters (3)

RLHF KL divergence penalty beta = not reported
The RLHF objective in §4.1 includes a fixed hyperparameter beta controlling the KL penalty to the base model; no value or tuning procedure is given.
Adversarial HK Value Bench question set size = 300
The benchmark uses 300 human-crafted questions (§5.3); the selection process and diversity analysis are not provided.
Sensitive-question test set size = 100
The sensitive question handling evaluation uses 100 self-constructed questions (§5.3); no release or validation is provided.

assumptions (4)

standard math Bradley-Terry preference model
Used in §4.1 to turn pairwise human preferences into a reward model loss; standard but not discussed for limitations.
domain assumption Local annotators' corrections represent Hong Kong values
The HKValue-Aligner is trained on a Q-A-C dataset from local annotators (§4.3); representativeness is explicitly acknowledged as uncertain in §6.1.
domain assumption Fine-tuning on local data preserves general capabilities
The claim that alignment efforts do not compromise core capabilities (§4.4) assumes no significant catastrophic forgetting, but no capability degradation analysis is provided.
ad hoc to paper Proprietary benchmarks are unbiased and contamination-free
HKMMLU, SafeLawBench, and NaVAB are cited from the authors' own prior work, and no contamination analysis is reported; this is needed to support the SOTA claims.

invented entities (2)

HKValue-Aligner
purpose: A correction model that refines base model responses toward Hong Kong values and generates preference data for RLHF (§4.3).
Not released, no external evaluation, and it is built on the existing Aligner method, so it provides no new falsifiable handle outside this paper.
Adversarial HK Value Benchmark
purpose: A proprietary evaluation tool with 300 human-crafted sensitive questions to quantify model alignment with Hong Kong ethical and legal standards (§5.3).
The benchmark is not released, not independently validated, and is authored by the same group, so its measurements are not externally checkable.

how reviews work

0 comments

Cite this review

Pith. "Pith review of HKGAI-V1: Towards Regional Sovereign Large Language Model for Hong Kong." pith.science (2026). https://pith.science/paper/5XMU27NK

@misc{pith2026250711502,
  author       = {Pith},
  title        = {Pith review of: HKGAI-V1: Towards Regional Sovereign Large Language Model for Hong Kong},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5XMU27NK}},
  note         = {Machine review of arXiv:2507.11502}
}

read the original abstract

This paper presents the development of HKGAI-V1, a foundational sovereign large language model (LLM), developed as part of an initiative to establish value-aligned AI infrastructure specifically tailored for Hong Kong. Addressing the region's unique multilingual environment (Cantonese, Mandarin, and English), its distinct socio-legal context under the "one country, two systems" framework, and specific local cultural and value considerations, the model is built upon the DeepSeek architecture and systematically aligned with regional norms through a multifaceted full parameter fine-tuning process. It is further integrated with a retrieval-augmented generation (RAG) system to ensure timely and factually grounded information access. The core contribution lies in the design and implementation of a comprehensive, region-specific AI alignment and safety framework, demonstrated through two key achievements: 1) The successful development of HKGAI-V1 itself - which outper-forms general-purpose models in handling Hong Kong-specific culturally sensitive queries, and embodies a "governance-embedded" approach to digital sovereignty - empowers Hong Kong to exercise control over AI applications in critical sectors including public services, legal systems, and edu-cation. 2) The development of the proprietary Adversarial HK Value Benchmark, a rigorous tool for evaluating model alignment with local ethical and legal stand-ards under challenging conditions. By documenting these achievements, the paper provides not only a technological artifact but also a replicable blueprint for developing advanced, regionally focused AI systems deeply rooted in their local identities.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

65 extracted references · 41 canonical work pages

[1]

Achiam, J., Adler, S., Agarwal, S., Ahmad, L., Akkaya, I., Aleman, F. L., ... & McGrew, B. (2023). Gpt-4 technical report. arXiv preprint arXiv:2303.08774

arXiv 2023
[2]

Claude 3

Anthropic. Claude 3. https://www.anthropic.com/news/claude-3-family, 2024a

work page
[3]

& Kaplan, J

Askell, A., Bai, Y., Chen, A., Drain, D., Ganguli, D., Henighan, T., ... & Kaplan, J. (2021). A general language assistant as a laboratory for alignment. arXiv preprint arXiv:2112.00861

arXiv 2021
[4]

& Summerfield, C

Bakker, M., Chadwick, M., Sheahan, H., Tessler, M., Campbell-Gillingham, L., Balaguer, J., ... & Summerfield, C. (2022). Fine-tuning language models to find agreement among humans with diverse preferences. Advances in Neural Information Processing Systems, 35, 38176-38189

work page 2022
[5]

Brandt, F., Conitzer, V., Endriss, U., Lang, J., & Procaccia, A. D. (Eds.). (2016). Handbook of computational social choice. Cambridge University Press

work page 2016
[6]

S., Schneider, J., Dragan, A., & Niekum, S

Brown, D. S., Schneider, J., Dragan, A., & Niekum, S. (2 021, July). Value alignment verification. In International Conference on Machine Learning (pp. 1105-1115). PMLR

work page
[7]

Measuring Hong Kong Massive Multi-Task Language Understanding

Cao, C., Zhu, Z., Zhu, J., Lu, G., Peng, S., Dai, J., ... & Guo, Y. (2025). Measuring Hong Kong Massive Multi - Task Language Understanding. arXiv preprint arXiv:2505.02177

work page Pith review arXiv 2025
[8]

Christiano, P., Shlegeris, B., & Amodei, D. (2018). Supervising strong learners by amplifying weak experts. arXiv preprint arXiv:1810.08575

arXiv 2018

Show all 65 references

[9]

R., Leibo, J

Dafoe, A., Hughes, E., Bachrach, Y., Collins, T., McKee, K. R., Leibo, J. Z., ... & Graepel, T. (2020). Open problems in cooperative AI. arXiv preprint arXiv:2012.08630

2020 arXiv
[10]

Dai, J., Chen, T., Wang, X., Yang, Z., Chen, T., Ji, J., & Yang, Y. (2024). Safesora: Towards safety alignment of text2video generation via a human preference dataset. Advances in Neural Information Processing Systems, 37, 17161-17214

2024
[11]

& Larson, J

Edge, D., Trinh, H., Cheng, N., Bradley, J., Chao, A., Mody, A., ... & Larson, J. (2024). From local to global: A graph rag approach to query -focused summarization. arXiv preprint arXiv:2404.16130

2024 arXiv
[12]

Gabriel, I. (2020). Artificial intelligence, values, and alignment. Minds and machines, 30(3), 411-437

2020
[13]

Gabriel, I., & Ghazavi, V. (2022). The challenge of value alignment. In The Oxford handbook of digital ethics. Oxfor d: Oxford University Press

2022
[14]

& Wang, H

Gao, Y., Xiong, Y., Gao, X., Jia, K., Pan, J., Bi, Y., ... & Wang, H. (2023). Retrieval -augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997, 2(1)

2023 arXiv
[15]

Gao, Z., Cao, Y., Wang, H., Ke, A., Feng, Y., Xie, X., & Zhou, S. K. (2025). FRAG: A Flexible Modular Framework for Retrieval-Augmented Generation based on Knowledge Graphs. arXiv preprint arXiv:2501.09957

2025 arXiv
[16]

I., Burnell, R., Bai, L., Gulati, A.,

Gemini Team, Georgiev, P., Lei, V. I., Burnell, R., Bai, L., Gulati, A., ... & Batsaikhan, B. O. (2024). Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context. arXiv preprint arXiv:2403.05530

2024 arXiv
[17]

Guo, D., Yang, D., Zhang, H., Song, J., Zhang, R., Xu, R., ... & He, Y. (2025). Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948

2025 arXiv
[18]

Hendrycks, D., Burns, C., Basart, S., Zou, A., Mazeika, M., Song, D., & Steinhardt, J. (2020). Measuring massive multitask language understanding. arXiv preprint arXiv:2009.03300

2020 arXiv
[19]

Digital Policy Office Releases Hong Kong Generative Artificial Intelligence Technical and Application Guideline

Hong Kong SAR Government. Digital Policy Office Releases Hong Kong Generative Artificial Intelligence Technical and Application Guideline. Info.gov.hk, 2025, www.info.gov.hk/gia/general/202504/15/P20250415 00227.htm. Accessed 26 May 2025

2025
[20]

https://github.com/PKU-Alignment/Beaver-zh-hk
[21]

& Song, Y

Hu, W., Li, H., Jing, H., Hu, Q., Zeng, Z., Han, S., ... & Song, Y. (2025). Context Reasoner: Incentivizing Reasoning Capability for Contextualized Privacy and Safety Compliance via Reinforcement Learning. arXiv preprint arXiv:2505.14585

2025 arXiv
[22]

& Lin, D

Huang, K., Liu, X., Guo, Q., Sun, T., Sun, J., Wang, Y., ... & Lin, D. (2023). Flames: Benchmarking value alignment of llms in chinese. arXiv preprint arXiv:2311.06899

2023 arXiv
[23]

& Chen, E

Huang, X., Liu, W., Chen, X., Wang, X., Wang, H., Lian, D., ... & Chen, E. (2024). Understanding the planning of LLM agents: A survey. arXiv preprint arXiv:2402.02716

2024 arXiv
[24]

& Yang, Y

Ji, J., Chen, B., Lou, H., Hong, D., Zhang, B., Pan, X., ... & Yang, Y. (2024). Aligner: Efficient alignment by learning to correct. Advances in Neural Information Processing Systems, 37, 90853-90890

2024
[25]

& Yang, Y

Ji, J., Chen, X., Pan, R., Zhu, H., Zhang, C., Li, J., ... & Yang, Y. (2025). Safe rlhf-v: Safe reinforcement learning from human feedback in multimodal large language models. arXiv preprint arXiv:2503.17682

2025 arXiv
[26]

& Yang, Y

Ji, J., Hong, D., Zhang, B., Chen, B., Dai, J., Zheng, B., ... & Yang, Y. (2024). Pku -saferlhf: Towards multi -level safety alignment for llms with human preference. arXiv preprint arXiv:2406.15513

2024 arXiv
[27]

& Yang, Y

Ji, J., Liu, M., Dai, J., Pan, X., Zhang, C., Bian, C., ... & Yang, Y. (2023). Beavertails: Towards improved safety 14 alignment of llm via a human -preference dataset. Advances in Neural Information Processing Systems, 36, 24678-24704

2023
[28]

& Gao, W

Ji, J., Qiu, T., Chen, B., Zhang, B., Lou, H., Wang, K., ... & Gao, W. (2023). Ai alignment: A comprehensive survey. arXiv preprint arXiv:2310.19852

2023 arXiv
[29]

& Yang, Y

Ji, J., Zhou, J., Lou, H., Chen, B., Hong, D., Wang, X., ... & Yang, Y. (2024). Align anything: Training all-modality models to follow instructions with language feedback. arXiv preprint arXiv:2412.15838

2024 arXiv
[30]

& Wen, J

Jin, J., Zhu, Y., Dou, Z., Dong, G., Yang, X., Zhang, C., ... & Wen, J. R. (2025, May). Flashrag: A modular toolkit for efficient retrieval-augmented generation research. In Companion Proceedings of the ACM on Web Conference 2025 (pp. 737-740)

2025
[32]

Klissarov, M., Hjelm, D., Toshev, A., & Mazoure, B. (20 24). On the Modeling Capabilities of Large Language M odels for Sequential Decision Making. arXiv preprint ar Xiv:2410.05656

arXiv
[33]

What is sovereign AI and why is it growing in importance?

Letort, Brian, and Kadri Linask -Goode. What is sovereign AI and why is it growing in importance? . Digitalrealty.com, 2025, www.digitalrealty.com/resources/articles/what-is- sovereign-ai. Accessed 26 May 2025

2025
[34]

& Kiela, D

Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., ... & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in neural information processing systems, 33, 9459-9474

2020
[35]

& S ong, Y

Li, H., Hu, W., Jing, H., Chen, Y., Hu, Q., Han, S., ... & S ong, Y. (2025). Privaci-bench: Evaluating privacy with c ontextual integrity and legal compliance. arXiv preprint arXiv:2502.17041

2025 arXiv
[36]

(2025, January)

Li, X. (2025, January). A Review of Prominent Paradigms for LLM -Based Agents: Tool Use, Planning (Including RAG), and Feedback Learning. In Proceedings of the 31st International Conference on Computational Linguistics (pp. 9760-9779)

2025
[37]

I., Liu, Z., Ferianc, M., Treleaven, P., & Rod rigues, M

Masoud, R. I., Liu, Z., Ferianc, M., Treleaven, P., & Rod rigues, M. (2023). Cultural Alignment in Large Langua ge Models: An Explanatory Analysis Based on Hofsted e's Cultural Dimensions. arXiv preprint arXiv:2309.1234 2

2023
[38]

Mügge, D. (2024). EU AI sovereignty: For whom, to what end, and to whose benefit?. Journal of European Public Policy, 31(8), 2200-2225

2024
[39]

Nicolson, K. (2016). Landscapes lost and found: appreciatin g Hong Kong’s heritage cultural landscapes . Hong Kong U niversity Press

2016
[40]

& Lowe, R

Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., ... & Lowe, R. (2022). Training language models to follow instructions with human feedback. Advances in neural information processing systems, 35, 27730-27744

2022
[41]

D., Ermon, S., & Finn, C

Rafailov, R., Sharma, A., Mitchell, E., Manning, C. D., Ermon, S., & Finn, C. (2023). Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36, 53728-53741

2023
[42]

Renze, M., & Guven, E. (2024). Self -reflection in llm agents: Effects on problem -solving performance. arXiv preprint arXiv:2405.06682

2024 arXiv
[43]

Russell, S. (2019). Human compatible: AI and the problem of control. Penguin Uk

2019
[44]

Scardovi, C., & Scardovi, C. (2021). From smart to meta cities. Sustainable Cities: Big Data, Artificial Intelligence a nd the Rise of Green,“Cy-phy” Cities, 1-20

2021
[45]

Scassa, T. (2023). Sovereignity and the governance of artificial intelligence. UCLA L. Rev. Discourse, 71, 214

2023
[46]

Sin, W. M. (2006). Law, autonomy and politics: The cha nging socio-political roles of law in postcolonial Hong Kong. International Journal of the Sociology of Law , 34(1), 64-83

2006
[47]

Singh, A., Ehtesham, A., Kumar, S., & Khoei, T. T. (2025). Agentic Retrieval -Augmented Generation: A Survey on Agentic RAG. arXiv preprint arXiv:2501.09136

2025 arXiv
[48]

One country, two systems

So, A. Y. (2011). “One country, two systems” and Hong Kong-China national integration: A crisis -transformati on perspective. Journal of Contemporary Asia , 41(1), 99-1 16

2011
[49]

& Naik, N

Wallace, B., Dang, M., Rafailov, R., Zhou, L., Lou, A., Purushwalkam, S., ... & Naik, N. (2024). Diffusion model alignment using direct preference optimization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 8228-8238)

2024
[50]

& Chen, W

Wang, Y., Ma, X., Zhang, G., Ni, Y., Chandra, A., Guo, S., ... & Chen, W. (2024, June). Mmlu-pro: A more robust and challenging multi -task language understanding benchmark. In The Thirty -eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track

2024
[51]

F., Michael, K., Pitt, J., & Evers, V

Winfield, A. F., Michael, K., Pitt, J., & Evers, V. (2019). Machine ethics: The design and governance of ethical AI and autonomous systems [scanning the issue]. Proceedings of the IEEE, 107(3), 509-517

2019
[52]

Xu, Y., Wang, F., An, Z., Wang, Q., & Zhang, Z. (2023). Artificial intelligence for science —bridging data to wis dom. The Innovation, 4(6)

2023
[53]

Yang, Y., Chai, H., Shao, S., Song, Y., Qi, S., Rui, R., & Zhang, W. (2025). Agentnet: Decentralized evolutionary coordination for llm -based multi -agent systems. arXiv preprint arXiv:2504.00587

2025 arXiv
[54]

(2021, December)

Yi Mak, H., & Lee, T. (2021, December). Low-resource n mt: A case study on the written and spoken languages i 15 n hong kong. In Proceedings of the 2021 5th International Conference on Natural Language Processing and Informatio n Retrieval (pp. 81-87)

2021
[55]

R., & Yang, Q

Yu, H., Shen, Z., Miao, C., Leung, C., Lesser, V. R., & Yang, Q. (2018). Building ethics into artificial intelligence. arXiv preprint arXiv:1812.02953

2018 arXiv
[56]

& Chua, T

Yu, T., Yao, Y., Zhang, H., He, T., Han, Y., Cui, G., ... & Chua, T. S. (2024). Rlhf -v: Towards trustworthy mllms via behavior alignment from fine -grained correctional human feedback. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 1...

2024
[57]

& Guo, H

Zhang, Y., Lin, Y., Zheng, G., Liu, Y., Sukiennik, N., Xu , F., ... & Guo, H. (2025). MetaCity: Data -driven sustain able development of complex cities. The Innovation

2025
[58]

Zhang, Y., Liu, H., Jiang, F., Luo, W., & Zhang, K. (2024 ). Building Decision Making Models Through Languag e Model Regime. arXiv preprint arXiv:2408.06087

2024 arXiv
[59]

Zhang, Y., Sun, R., Chen, Y., Pfister, T., Zhang, R., & Arik, S. (2024). Chain of agents: Large language models collaborating on long -context tasks. Advances in Neural Information Processing Systems, 37, 132208-132237

2024
[60]

Zheng, X., Weng, Z., Lyu, Y., Jiang, L., Xue, H., Ren, B., ... & Hu, X. (2025). Retrieval augmented generation and understanding in vision: A survey and new outlook. arXiv preprint arXiv:2503.18016

2025 arXiv
[61]

& Liu, T

Zhong, T., Yang, Z., Liu, Z., Zhang, R., Liu, Y., Sun, H., ... & Liu, T. (2024). Opportunities and challenges of larg e language models for low-resource languages in huma nities research. arXiv preprint arXiv:2412.04497

2024 arXiv
[62]

& Duan, N

Zhong, W., Cui, R., Guo, Y., Liang, Y., Lu, S., Wang, Y., ... & Duan, N. (2023). Agieval: A human -centric benchmark for evaluating foundation models. arXiv preprint arXiv:2304.06364

2023 arXiv
[63]

(2025, April)

Zhou, J., Ji, J., Dai, J., & Yang, Y. (2025, April). Sequence to sequence reward modeling: Improving rlhf by language feedback. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 39, No. 26, pp. 27765 - 27773)

2025
[64]

& Xio ng, D

Zhu, S., Xu, S., Sun, H., Pan, L., Cui, M., Du, J., ... & Xio ng, D. (2024). Multilingual Large Language Models: A Systematic Survey. arXiv preprint arXiv:2411.11072

2024 arXiv
[65]

& Gu o, Y

Cao, C., Zhu, H., Ji, J., Sun, Q., Zhu, Z., Wu, Y., ... & Gu o, Y. (2025). SafeLawBench: Towards Safe Alignment of Large Language Models. arXiv preprint arXiv:2506.0663 6

2025
[66]

& G uo, Y

Shi, W., Ju, C., Liu, C., Ji, J., Zhang, J., Zhang, R., ... & G uo, Y. (2025). Benchmarking Multi -National Value Alig nment for Large Language Models. arXiv preprint arXiv :2504.12911

2025 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Achiam, J., Adler, S., Agarwal, S., Ahmad, L., Akkaya, I., Aleman, F. L., ... & McGrew, B. (2023). Gpt-4 technical report. arXiv preprint arXiv:2303.08774

arXiv 2023

[2] [2]

Claude 3

Anthropic. Claude 3. https://www.anthropic.com/news/claude-3-family, 2024a

work page

[3] [3]

& Kaplan, J

Askell, A., Bai, Y., Chen, A., Drain, D., Ganguli, D., Henighan, T., ... & Kaplan, J. (2021). A general language assistant as a laboratory for alignment. arXiv preprint arXiv:2112.00861

arXiv 2021

[4] [4]

& Summerfield, C

Bakker, M., Chadwick, M., Sheahan, H., Tessler, M., Campbell-Gillingham, L., Balaguer, J., ... & Summerfield, C. (2022). Fine-tuning language models to find agreement among humans with diverse preferences. Advances in Neural Information Processing Systems, 35, 38176-38189

work page 2022

[5] [5]

Brandt, F., Conitzer, V., Endriss, U., Lang, J., & Procaccia, A. D. (Eds.). (2016). Handbook of computational social choice. Cambridge University Press

work page 2016

[6] [6]

S., Schneider, J., Dragan, A., & Niekum, S

Brown, D. S., Schneider, J., Dragan, A., & Niekum, S. (2 021, July). Value alignment verification. In International Conference on Machine Learning (pp. 1105-1115). PMLR

work page

[7] [7]

Measuring Hong Kong Massive Multi-Task Language Understanding

Cao, C., Zhu, Z., Zhu, J., Lu, G., Peng, S., Dai, J., ... & Guo, Y. (2025). Measuring Hong Kong Massive Multi - Task Language Understanding. arXiv preprint arXiv:2505.02177

work page Pith review arXiv 2025

[8] [8]

Christiano, P., Shlegeris, B., & Amodei, D. (2018). Supervising strong learners by amplifying weak experts. arXiv preprint arXiv:1810.08575

arXiv 2018