A well-executed PhD full of negative results should still be defendable
prajdabre · x · 2026-08-04
The author argues that a PhD thesis full of well-executed negative results should be defendable, because the field heavily rewards positive findings and SoTA claims while undervaluing failed attempts.
They point to two workshops as examples of efforts to normalize this kind of work: “I can’t believe it’s not better” and “Insights from Negative Results in NLP.” The core message is that negative results are still real results and should be reported more openly.
Related event: Researchers Call for Embracing Negative Results in AI(2 posts)→
More from Research
- AFM3 20B paper proposes prompt-level pruning that activates only 20% of MLP layers — Aaaaaaaaaeeeee · 2026-08-04
- This week’s AI paper roundup spans video world models, red-teaming, memory and search — TheTuringPost · 2026-08-04
- Universal Transformers revisit recurrent bias, dynamic halting and Turing completeness — agihippo · 2026-08-04
- New Benchmark Probes Transformer Generalization Over Task Complexity — jasondeanlee · 2026-08-04
- Microsoft Research shows weak models can still distill stronger students — MicrosoftResearch · 2026-08-04
- AI Researcher Compares Hyper-UT and Tilde: Squaring Modulo Problem Better Tests Universal Transformers — kohjingyu · 2026-08-04