Hao Li

Shanghai Jiao Tong University · China

Unclaimed · AI-structured profileOpenAlex

Hao Li — Shanghai Jiao Tong University researcher bridging efficient LLM architecture and quantum information science

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-14) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Output is not yet scored: this OpenAlex record covers only one year (3 works in 2026). This axis would measure total publications, yearly cadence, and authorship positions across venues, all unavailable here.

Impact—

Impact is not yet scored: OpenAlex reports h-index and i10-index of 0 with no citation aggregate. This axis would measure citation counts, h-index trajectory, and influence on downstream work; with a record this thin, no impact estimate is attempted.

Focus0/100

Focus scores 0/100 because OpenAlex splits this profile exactly 50/50 between two distant fields — efficient large language model architecture and quantum information science — and the topic list likewise spans quantum information and cryptography (2 works), topic modeling, natural language processing, advanced neural network applications, and quantum mechanics. The pattern indicates a deliberate cross-domain spread rather than concentration in a single research lane.

Collaboration85/100

Collaboration scores 85/100, backed by 70 distinct coauthors on the OpenAlex record — an unusually large pool for only 3 recorded works, consistent with the big-team authorship model of the DeepSeek-V4.1-Flash paper, which lists DeepSeek-AI as an institutional coauthor. The quantum state characterization paper adds a second, more conventional collaboration cluster, suggesting comfort working inside both consortium-scale and mid-size laboratory teams.

Activity40/100

Activity scores 40/100: OpenAlex countsByYear lists a single year, 2026, with 3 works, and recent3yWorks is likewise 3. All visible output is concentrated in one recent year with no longer publication history on record, so the mid-range score reflects genuine but very recent and shallow-by-volume activity rather than a sustained multi-year cadence.

Reproducibility38/100

Reproducibility scores 38/100, averaged over 2 Evolve-reviewed papers (reproMean 38). The DeepSeek-V4.1-Flash paper rates low: Evolve notes the 552B-model results are self-evaluated, without ablations, and Table 1 results are mixed. The quantum remote-state-preparation paper rates medium, with the >99.999% fidelity reconstruction experiment judged strong but limited to a narrow state family.

Research summary

Hao Li is a researcher at Shanghai Jiao Tong University whose publicly indexed output spans two distinct fields in near-equal measure: OpenAlex assigns a 50/50 field distribution between efficient large language model architecture and quantum information science. The Evolve catalog holds two 2026 arXiv papers. The first, DeepSeek-V4.1-Flash, where Li is last author, reports a 552B-parameter mixture-of-experts model claimed to match frontier agents at one-quarter to one-eighth the KV cache through sparse attention, FP4 quantization, and speculative decoding; Evolve judges it engineering-strong but self-evaluated, with no ablations and mixed Table 1 results, and rates reproducibility readiness low. The second, on noise-robust quantum state characterization for remote state preparation, places Li in a middle authorship position and applies a transformer denoiser to reconstruct noisy photonic states above 99.999% fidelity — a strong experiment, though on a narrow state family with marginal gains over a matched MLP. Bibliometrically, OpenAlex records 3 works in the last three years (all 2026), 70 distinct coauthors, an h-index of 0, and no citation-impact score yet; radar scores are collaboration 85, activity 40, focus 0, reproducibility 38. As 'Hao Li' is a very common name and no ORCID is linked, this profile is name-resolved from public bibliographic data.

Collaboration pattern

The OpenAlex record lists 70 distinct coauthors across just 3 works — a ratio that points to large-team, consortium-style papers, exemplified by DeepSeek-AI appearing as an institutional coauthor on the KV cache compression work. The frequent-coauthor list is dominated by romanized Chinese names (e.g., B. Li, Bochao Wu, Bowei Zhang), consistent with a Shanghai Jiao Tong University base. One parsed entry is the string ':', indicating incomplete author-name disambiguation in the source data.

Research keywords

KV cache compressionsparse attentionmixture-of-expertsFP4 quantizationspeculative decodingremote state preparationquantum state tomographyTransformer denoisingdensity matrix reconstructionphotonic polarization qubits

Highlights

DeepSeek-V4.1-Flash: KV cache compression at frontier-agent scale
As last author, Li contributed to a 552B-parameter mixture-of-experts model claimed to match frontier agents at 1/4-1/8 of the KV cache using sparse attention, encoder-decoder design, FP4 quantization, and speculative decoding. Evolve judges the engineering strong but the evaluation self-reported, with no ablations and mixed Table 1 results; reproducibility readiness is low.
Transformer denoising for noise-robust quantum state characterization
In a middle authorship position, Li co-authored work applying a transformer denoiser to remote state preparation, reconstructing noisy photonic polarization-qubit states and density matrices at >99.999% fidelity. Evolve rates the experiment strong but notes the narrow state family and marginal gains over a matched MLP; reproducibility readiness is medium.

Timeline

2026OpenAlex records 3 works in 2026 — the only publication year on this profile, so the visible record begins here.
2026Last-author paper on DeepSeek-V4.1-Flash pushes KV cache compression for a 552B mixture-of-experts model.
2026Middle-author paper applies transformer denoising to noise-robust quantum state characterization for remote state preparation.

Papers in the Evolve catalog

Frequent co-authors

DeepSeek-AI×1 :×1 Anyi Xu×1 B. Li×1 Bangcai Lin×1 Bing Xue×1 BingCheng Xian×1 Bingzheng Xu×1 Bochao Wu×1 Bowei Zhang×1 Boyi Deng×1 C. C. Yu×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

All bibliometrics come from OpenAlex; h-index and i10-index are recorded as 0 and the output/impact radar scores are missing, so citation-based impact cannot yet be assessed.
'Hao Li' is a very common name and no ORCID is linked; this profile is name-resolved from public bibliographic data and may mix records of different researchers.
Coauthor strings include non-person entries ('DeepSeek-AI' and ':'), indicating incomplete author parsing in the source record.
Qualitative assessments rest on only two Evolve-reviewed papers, both 2026 arXiv preprints; no peer-reviewed venue information is on record.