Evolve.Pub
Sign inScan your paper
KC

Kuanlin Chen

National Tsing Hua University · Taiwan

Unclaimed · AI-structured profileOpenAlex

Applies reinforcement learning and 4-bit quantized LLMs to battery storage dispatch and edge AI information extraction

3

papers here

3

reviewed by Evolve

3

first-author

0

corresponding

—

works · OpenAlex

0

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-09) 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: insufficient dataOutputImpact: insufficient dataImpactFocus: 0/100FocusCollaboration: 22/100CollaborationActivity: 40/100ActivityReproducibility: 55/100Reproducibility

Output—

Output cannot be scored yet: the snapshot records just 3 recent works with h-index 0 and i10-index 0, insufficient for a lifetime productivity estimate; this axis would measure publication volume and its growth over time.

Impact—

Impact is not scoreable: with 0 citations on record and all three papers posted to arXiv in 2026, there is no citation base; this axis would measure citation counts, h-index trajectory, and uptake by other researchers.

Focus0/100

The focus score is 0, reflecting an unusually even spread: the field distribution splits exactly three ways (0.333 each across RL for power systems, LLM efficiency/quantization, and NLP/edge AI), and OpenAlex topics scatter across five clusters — Topic Modeling and Advanced Neural Network Applications (2 each) plus NLP techniques, microgrid control, and power system optimization (1 each). The through-line is applied ML under deployment constraints, but the portfolio is still too young to have narrowed.

Collaboration22/100

Collaboration scores 22, near the low end: the snapshot records only 2 distinct co-authors across the 3 catalogued papers, both recurring — Chen-Wei Kuo and Cheng-En Ou. All three papers are first-authored by Chen with no corresponding-author credit recorded, consistent with a small-team or largely independent working mode. The base is thin: this pattern reflects an early, compact collaboration network rather than a demonstrated preference for solo work, and should be re-read as more co-authored output appears.

Activity40/100

Activity scores 40, mid-range for an early-stage record: all 3 works in the snapshot fall in 2026 (countsByYear lists 2026: 3, and recent3yWorks equals 3), i.e. three arXiv preprints appearing within a single year. That burst shows a currently active submission cadence, but with no earlier years on record the baseline is a single data point — the score can neither confirm sustained throughput nor rule out one concentrated push.

Reproducibility55/100

Reproducibility scores 55, a mean across all 3 catalogued papers, each rated medium readiness. Evolve's reviews repeatedly cite concrete, checkable numbers — Spearman ρs ≥ 0.980 and top-k ranges of 0.619–0.793 in the quantization study; 82.9% F1, 5.3 s P50, and 12.5 W for ARID — while flagging thin statistics, thin baselines, and no variance data. Reporting is candid and quantified, but not yet at a level that guarantees independent replication.

Research summary

Kuanlin Chen is a researcher at National Tsing Hua University whose 2026 output, catalogued by Evolve from arXiv preprints, spans three applied threads: cost-aware reinforcement learning for battery energy storage dispatch, 4-bit weight-only quantization of large language models, and deployable edge AI for industrial information extraction. All three catalogued papers list Chen as first author. The OpenAlex-derived snapshot records 3 works in the most recent three-year window with an h-index of 0 and an i10-index of 0, consistent with a very early-stage record whose citation impact cannot yet be scored. The storage study combines action masking and projection with advisory PPO for electricity market arbitrage under suppressed spreads; Evolve's review finds the claims honest but notes PPO underperforms model predictive control and profits are trivially small. The quantization study shows NF4 preserves global BOS-sink head ranks (ρs ≥ 0.980) while top-k head sets shift (0.619–0.793) across three small LLMs. The ARID system distills a 4-bit LLM with QLoRA and pseudo-labeling to extract 4-field JSON from maintenance work orders on a Jetson Orin NX, reporting 82.9% F1, 5.3 s P50 latency, and 12.5 W. Frequent co-authors are Chen-Wei Kuo and Cheng-En Ou (2 distinct co-authors recorded).

Collaboration pattern

The recorded network is compact: 2 distinct co-authors, both recurring across the 2026 preprints — Chen-Wei Kuo and Cheng-En Ou. Chen takes first authorship on all three catalogued papers, with no corresponding-author role recorded, suggesting a small stable team where Chen leads experiments and writing. No senior-lab or cross-institution collaborators appear in the snapshot, so the collaboration score of 22 mainly reflects sample size; expect this picture to change quickly with further output.

Research keywords

reinforcement learningbattery energy storage dispatchelectricity market arbitragesafe reinforcement learningaction maskingpost-training quantizationNF4 weight-only quantizationattention sinkedge AI deploymentinformation extractionLLM distillationQLoRA fine-tuning

Highlights

Timeline

  1. 2026Entire catalogued record falls in 2026: three first-author arXiv preprints, with the snapshot recording an h-index of 0 and 3 works in the last three years.
  2. 2026First catalogued preprint: cost-aware reinforcement learning with action masking and projection for battery energy storage dispatch under suppressed-spread market shifts.
  3. 2026Second preprint: empirical study of BOS-sink attention topology under 4-bit weight-only quantization across three small LLMs.
  4. 2026Third preprint: ARID, a deployable edge AI system extracting structured fields from industrial maintenance work orders on Jetson hardware.

Papers in the Evolve catalog

Frequent co-authors

Chen-Wei Kuo×3Cheng-En Ou×2

Is this you? Claim this profile

Enter the email you published with. We'll send a one-time claim link; after claiming you can edit the introduction or hide this page. Requests from a mismatched address go to manual review.

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.

  • Bibliometrics (h-index, i10-index, topics, work and co-author counts) come from an OpenAlex-derived snapshot only; the output and impact axes are unscored because citation data is absent.
  • No ORCID was available, so the profile is name-resolved from public bibliographic data; Kuanlin Chen is a common name and identity should be confirmed before citing.
  • All quoted results are as summarized in Evolve's reviews of arXiv preprints and are not peer-review validated; all three papers are rated medium reproducibility readiness.
  • The record contains only 2026 works, so no multi-year trend is computable, and with only 2 distinct co-authors the collaboration findings are thin.
Kuanlin Chen — Researcher Profile | Evolve.Pub