Skip to content
View ashioyajotham's full-sized avatar
💭
Data! Data! Data!
💭
Data! Data! Data!
97 followers · 50 following

Achievements

Achievement: StarstruckAchievement: Pair ExtraordinaireAchievement: YOLOAchievement: Pull Sharkx2Achievement: Quickdraw

Achievements

Achievement: StarstruckAchievement: Pair ExtraordinaireAchievement: YOLOAchievement: Pull Sharkx2Achievement: Quickdraw

Highlights

Block or report ashioyajotham

Report abuse

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

Report abuse
ashioyajotham/README.md

🔭 Currently Working On

  • Building assistants for information understanding
  • Technical AI Safety
  • Machine learning systems design

🌱 Research Interests

  • AI Safety (red-teaming aka jailbreaking, AI control, mech interp and alignment)
  • Foundation Models (pretraining, post-training, confabulations ["hallucinations"], scaling laws)
  • World models (physical AI, spatial intelligence etc)

📝 Latest Blog Posts

▶ Read more on Substack

GitHub followers GitHub User's stars

Typing SVG

📈 GitHub Analytics

GitHub Streak

⚡ Recent GitHub Activity
Ashioya's Activity Graph

Pinned Loading

  1. fingpt_trader fingpt_trader Public template

    A quant trading system platform based on FinGPT, demonstrating new applications of large pre-trained Language Models in quantitative finance.

    Python 27 8

  2. cot-faithfulness-mech-interp cot-faithfulness-mech-interp Public

    I am investigating the mechanistic architecture of unfaithful Chain-of-Thought (CoT), specifically mapping and disrupting the “shortcut circuits” that allow models to bypass explicit reasoning.

    Jupyter Notebook 4

  3. greater-than-circuit greater-than-circuit Public

    Reverse engineering the circuit responsible for the "greater than" capability in a language model

    HTML

  4. weather_forecasting_lora weather_forecasting_lora Public

    How close can LoRA get to full fine-tuning (FullFT) in terms of learning speed, performance, and compute tradeoffs? And under what conditions?

    Python

  5. spot-the-flaw spot-the-flaw Public

    Verifier-backed math reasoning lab for proof flaw localization, minimal repair, and RL-style evaluation.

    Python

  6. global-activity-monitor global-activity-monitor Public

    Autonomous geopolitical situation tracker -- discovers, scores, and maps active global events in real time using dual GDELT APIs, TF-IDF summarisation, and SQLite persistence

    JavaScript 3