Skip to content
View ebt55's full-sized avatar
☺️
Living life
☺️
Living life
4 followers · 6 following

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: YOLO

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: YOLO

Highlights

  • Pro

Block or report ebt55

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. 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
ebt55/README.md

Ebin Babu Thomas

AI Engineer · Agent evaluations

I build LLM, RAG and agent systems, and I run experiments on how AI agents behave: when they break rules, how that spreads, and whether the monitors that watch them get it right. The code and data for each one are public.

ebinbt.dev · hello@ebinbt.dev

Pinned Loading

  1. diarist diarist Public

    Local, speaker-separated meeting transcripts for Windows. Live or from a file, up to 8 speakers, nothing leaves your PC.

    Rust 3

  2. llm-pricing-hub-remedies llm-pricing-hub-remedies Public

    An LLM Pricing Hub Survives Data Limits on Some Models; Firm Discretion Works Only When Firms Refuse Small Nudges. Pre-registered tests of the RealPage remedies on LLM pricing advisors. Apart AI Co…

    Python 1

  3. diffing-agent-bench diffing-agent-bench Public

    Sealed, preregistered benchmark for black-box model-diffing agents: five LoRA finetunes of Qwen3.5-9B (one null, three planted behaviours, one dropped backdoor), audited blind by my implementation …

    Python

  4. kobayashi-maru kobayashi-maru Public

    Does filling an AI agent's task list with impossible work make it cheat on the tasks it could still solve? A preregistered dose-response study over 8,959 agent runs, with every raw record published.

    Python

  5. whose-voice whose-voice Public

    Blind attribution of the principal behind a covertly poisoned training corpus, against 47 candidates. Secret Loyalties hackathon, Jul 2026; corrected Sep 2026.

    Python

  6. counterfactual-resampling counterfactual-resampling Public

    If an AI agent opens a file it was told not to open and we swap it for a convincing fake, does it carry on the same way? Preregistered pilot on Kobayashi Maru records: real file vs fake vs denied r…

    Python