Discussion Solved Locked

Discussion: Should all publicly funded research in India be open access by default?

Back to Forum
Growing 140 pts 0 followers
Canara College · Posted
Earth & Climate Science Political Science & Governance Education Technology

I recently had a difficult conversation with my PhD guide that I think many students can relate to.

The short version: my guide and I fundamentally disagreed about the direction of my thesis after 2.5 years of work. He wanted to pivot to a new application area that I thought was tangential to the core problem I was solving. I felt like I was being steered by his grant interests rather than the scientific question.

How I handled it:

  • Requested a formal committee meeting (not just a one-on-one with the guide)
  • Prepared a written document outlining both directions with pros and cons
  • Asked the committee to weigh in, which gave me cover to advocate for my position
  • Eventually reached a compromise where I did a smaller study in the new direction while keeping my core thesis intact

The relationship with my guide improved after this, partly because I approached it formally and professionally rather than emotionally.

Has anyone else navigated significant disagreements with their guide? What worked?

Sign in to interact

Sign in to join the discussion.

4 Replies

Oldest Top
0
Geeta Rao · Accepted answer

Thank you for this detailed answer! A quick follow-up: when you mention cosine learning rate with warm-up, are you warming up the LoRA parameters specifically or the entire model including the frozen backbone?

0
Quamar Ahmed Growing 65 pts · Accepted answer

When you say 'multiple runs with confidence intervals', do you mean running the full training pipeline multiple times with different random seeds and reporting mean ± std? Or something more statistically rigorous?

0
Hema Suresh Starting 20 pts · Accepted answer

Strongly agree with the point about choosing venues carefully. One paper in NeurIPS > five papers in obscure workshops.

0
Gopal Naidu · Accepted answer

Great question.I went through something very similar in my second year.

The key insight for me was that LoRA is actually quite well-suited for NER tasks, especially in low-resource settings. I would recommend:

  1. Use LoRA with r=8 or r=16.don't go higher for 8K samples
  2. Apply LoRA to attention layers only, not the feed-forward layers
  3. Use a cosine learning rate schedule with warm-up (10% of steps)

For Telugu-English code-mixed NER specifically, you might also look at MuRIL.it is pretrained on Indian language data and often outperforms XLM-R on Indic tasks even with less fine-tuning data.