Blog

Research notes, engineering logs, reading group takeaways, and the occasional reflection.

Posts

Why Critique Signals Beat Scalar Rewards for Tree Search

Reflections on the core insight behind CATPO — why structured natural language critiques at each tree node provide fundamentally better optimization gradients than scalar reward signals from the leaf. Drawing parallels to how humans debug code: we don't just rate the output, we reason about intermediate steps…

research · rl · alignment

Lessons From Optimizing Multi-Node Training Pipelines

How we reduced wall-clock time per epoch by 23% on our shared GPU cluster. This post covers the unglamorous but important stuff: gradient accumulation tuning, mixed-precision scheduling quirks, DataLoader bottlenecks, and why your training is secretly I/O bound…

infrastructure · gpu · systems

VLG Reading Group: Scaling Laws for Pruned Subnetworks

Notes from our VLG internal presentation on how Chinchilla-style scaling laws interact with evolutionary pruning. If you plan to prune a model to 50% sparsity, should you train a larger model with less data, or a smaller model with more data?

reading group · scaling · pruning

Hunting for IOI Circuits in Small Transformers

Replicating Anthropic's indirect object identification circuit discovery on GPT-2 Small and Pythia 410M. What survives across model families, what doesn't, and what that tells us about the universality of learned circuits…

interpretability · circuits · transformers

Dispatch Log

Short-form updates — what I shipped, read, presented, or learned.

Jun 2, 2026

Submitted CATPO to COLM 2026.

Jun 1, 2026

Pipeline optimization pass complete — 23% wall-clock reduction on multi-node cluster.

Jun 1, 2026

Deployed automated GPU health-check scripts across VLG compute pool.

May 22, 2026

Presented evolutionary pruning framework at VLG reading group.

May 10, 2026

CATPO shows 12-point win-rate improvement vs. DPO with 40% fewer rollouts.

Apr 15, 2026

"Natural Selection of LLMs via Pruning" accepted at AAAI 2026 Student Abstract.

Apr 5, 2026

Finalized CATPO prototype — evaluation pipeline + AlpacaEval-2 benchmarks wired up.

Mar 18, 2026

Started mech interp experiments — IOI circuit analysis on GPT-2 Small & Pythia 410M.

Mar 5, 2026

Set up TransformerLens + nnsight for activation patching experiments.

Feb 12, 2026

Joined VLG at IIT Roorkee as a core research member.

Jan 8, 2026

Started B.Tech in Data Science & AI at IIT Roorkee.