Dolby researcher bridging generative audio mixing and multimodal visual reasoning with renderable programs
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-02) 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—
No publication-volume data is available for this profile yet, so output is unscored. This axis would measure the number and cadence of peer-reviewed papers and preprints over time, which cannot be derived from the two reviewed catalog entries alone.
Impact—
Impact is unscored because citation counts, h-index, and usage signals are absent from the current data. This axis would measure scholarly uptake — citations of the reviewed papers and downstream adoption of methods like generative audio mixing and SVG-based visual reasoning.
Focus0/100
Focus scores 0, and the raw field distribution explains why: activity splits exactly 0.5/0.5 between audio signal processing / machine learning and machine learning (multimodal vision-language models), based on the two reviewed 2026 papers. This is an even two-field spread rather than one dominant line. Whether it reflects deliberate breadth or an early-stage career still settling cannot be judged from two catalog entries.
Collaboration54/100
Collaboration scores 54, drawn from 14 distinct coauthors across the reviewed work. The pattern is hybrid: Dolby-side collaborators such as Ilpo Viertola, Giulio Cengarle, Gouthaman KV, and Daniel Arteaga anchor the audio mixing paper, while the vision-language paper connects to academic groups including Joyce Chai, Chuang Gan, and Hao Zhang. Lu sits mid-authorship on one paper and last author on the other, neither marked corresponding.
Activity—
Activity is unscored: the data provides no publication timeline beyond two 2026 preprints, so recent-output velocity cannot be computed. This axis would measure year-over-year publishing rate and sustained engagement, requiring a fuller bibliometric record than the current catalog offers.
Reproducibility38/100
Reproducibility scores 38, the mean across 2 scored papers. Evolve's review of the SVGLM paper notes no significance tests or code release are stated, and the audio mixing paper's readiness is rated low (the vision-language paper, medium). Strong reported gains — about 6.5-10 weighted points over SFT — therefore remain hard to verify independently until releases accompany results.
Research summary
Lie Lu is a researcher at Dolby (Netherlands) whose Evolve-reviewed output to date spans two evenly weighted strands: audio signal processing with machine learning, and multimodal vision-language modeling — the field distribution assigns 0.5 to each. On the audio side, Lu is last author of 'Spot, Separate, and Enhance' (2026), a fully generative audio mixing system built on conditional flow matching and a diffusion transformer that remixes user-generated content from text and video cues; Evolve's verdict finds it beats the VisAH baseline on generative metrics and human preference while trailing VisAH-FM on reconstruction. On the vision-language side, Lu contributes as a middle author to 'Multimodal Thinking with Renderable Programs' (2026), where SVGLM tunes vision-language models to sketch SVG overlays while reasoning, gaining roughly 6.5-10 weighted points over SFT on geometry — though Evolve notes no significance tests or code release are stated. The radar places collaboration at 54 (14 distinct coauthors spanning Dolby colleagues and academic groups), focus at 0 (the exact two-field split), and reproducibility at 38 (the mean over 2 scored papers). Output, impact, and activity metrics are unavailable, making this an early-stage, name-resolved profile built from public bibliographic data and Evolve's paper reviews.
Collaboration pattern
Lu's collaboration pattern mixes industry and academia: 14 distinct coauthors appear across the reviewed work, with a Dolby-side cluster (Ilpo Viertola, Giulio Cengarle, Gouthaman KV, Daniel Arteaga) tied to the audio mixing paper and academic ML collaborators (Sunli Chen, Ding Zhong, Ziqiao Ma, Jiaxin Liu, Zeyuan Yang, Hao Zhang, Joyce Chai, Chuang Gan) aligning with the vision-language paper. Recorded authorship sits at middle and last position, with no corresponding-author credit.
Research keywords
Highlights
SVG overlays as a reasoning medium for vision-language models
Fully generative, text-and-video-guided audio mixing
Cross-domain span with an openness gap
Within a single year, Lu's two reviewed 2026 preprints cover both audio and multimodal reasoning, yet reproducibility scores 38/100 across 2 papers and release readiness is low-to-medium — both papers' gains currently rest on reported numbers without stated significance testing or artifacts.
Timeline
- 2026Last author on 'Spot, Separate, and Enhance', a fully generative text-and-video-guided audio mixing approach released as an arXiv preprint.
- 2026Middle author on 'Multimodal Thinking with Renderable Programs', introducing SVGLM for SVG-based visual reasoning as an arXiv preprint.
- 2026Both reviewed papers appear in the same year across audio and vision-language research, forming a dual-field profile supported by 14 distinct coauthors.
Papers in the Evolve catalog
Frequent co-authors
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.
- Name resolution: no ORCID or country is on record, and 'Lie Lu' is a common name; this profile is resolved from public bibliographic data and Evolve's reviewed papers.
- Bibliometric dimensions are derived from public bibliometric aggregation (OpenAlex-style sources) only; the reviewed evidence covers just two 2026 arXiv preprints with no peer-reviewed venue recorded.
- Output, impact, and activity axes are unscored because citation counts and publication-volume data are not yet available for this researcher.
- The pairing of frequent coauthors with specific papers is inferred from institutional affiliation and Evolve's verdicts, not from confirmed per-paper author lists.