Skip to content
View edward-lcl's full-sized avatar
0 followers · 6 following

Achievements

Achievement: Public SponsorAchievement: Pull Sharkx2Achievement: Pair Extraordinairex2Achievement: YOLOAchievement: Quickdraw

Achievements

Achievement: Public SponsorAchievement: Pull Sharkx2Achievement: Pair Extraordinairex2Achievement: YOLOAchievement: Quickdraw

Highlights

  • Pro

Block or report edward-lcl

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
edward-lcl/README.md

Edward Lue Chee Lip

evaluating agents · unsupervised, unfortunately.

I study how to evaluate and supervise AI agents, and build tools that make experiments easier to inspect and reproduce.

Selected work

  • Terminal-Bench: hardness and verifier audit — distinguishing genuine task difficulty from task defects and evaluation failures. Poster accepted at the NeurIPS 2026 Verify-Agents workshop; submitted to AAMAS 2027, not accepted. Related note.
  • Factor(U,T) — testing what a trusted monitor can detect from an untrusted model’s plan. Accepted at the AAAI 2026 TrustAgent workshop. Paper · Digest.
  • Eval Evidence — evaluation records with configuration, artifacts, and file hashes. Reproducible walkthrough.

Recent notes

Website · All research & projects · CV · Email · LinkedIn · X

Pinned Loading

  1. factor-ut-untrusted-decomposer factor-ut-untrusted-decomposer Public

    Factor(UT): Controlling Untrusted AI Decomposers — AAAI 2026 workshop paper on monitoring untrusted decomposition in code generation workflows.

    Jupyter Notebook 3 2

  2. edward-lcl.github.io edward-lcl.github.io Public

    Personal website — high-leverage systems, research, and cross-domain project history.

    TypeScript

  3. DeceptionScoop DeceptionScoop Public

    Side-channel profiler that detects deceptive intent in LLMs by measuring the computational cost of lying.

    Python

  4. bcb-deep-equal bcb-deep-equal Public archive

    Semantic deep-equality for BigCodeBench floating-point outputs. Published on PyPI.

    Python