Skip to content
View ydnyshhh's full-sized avatar

Highlights

  • Pro

Block or report ydnyshhh

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.

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
ydnyshhh/README.md

Project banner

Hi Yadnyesh Here! I am an aspiring ML researcher interested in both the foundations and the frontiers of AI. My work is motivated by questions around reasoning, reinforcement learning, post-training, ML systems, agentic systems, mechanistic interpretability, alignment, scientific machine learning and large foundation models. I am particularly excited by the challenge of designing AI systems that are more interpretable, efficient and scalable, especially when they can contribute to scientific progress and meaningful practical applications.

Pinned Loading

  1. ydnyshhh.github.io ydnyshhh.github.io Public

    HTML

  2. rlvr-gym rlvr-gym Public

    RLVR-Gym is a procedural environment generation library for verifiable reasoning and decision tasks, designed for RLVR, post-training research and structured evaluation. It generates formal, reprod…

    Python

  3. rewardhack-gym rewardhack-gym Public

    rewardhack-gym is a gym-style package for constructing and analyzing verifiable agent environments with controllable proxy-objective mismatch, designed for reward hacking, post-training and mechani…

    Python

  4. synthetic-workspace-gym synthetic-workspace-gym Public

    Synthetic Workspace Gym is a modular framework for generating executable synthetic workspace environments for agents, with hidden evaluators, controllable difficulty and full trajectory logging. It…

    Python

  5. trace2eval trace2eval Public

    Trace2Eval is a local-first Python toolkit for turning failed long-horizon coding-agent runs into small, executable regression evals.

    Python

  6. moral-mechinterp moral-mechinterp Public

    moral-mechinterp is a reproducible research repo for studying how reward-adapted Qwen agents behave on the GT-HarmBench dataset, and where those behavioral differences appear inside the model.

    Python