Tag
The paper demonstrates that prompt-space meta-learning for personalizing frozen large language models does not transfer across users, as the meta-validation objective is statistically invariant to user-support correspondence, leading to no significant improvement over seed prompts or controls.
This paper presents a negative result showing that current knowledge-editing benchmarks cannot effectively evaluate scope classifiers, using INLAY, a gradient-free editor, to demonstrate that no per-query routing method can improve performance due to structural limitations in the benchmarks.
This paper presents a pre-registered negative result on training mixture-of-experts routers for cache locality against memory-bandwidth walls, showing that miss reduction trades off with language modeling quality despite training mechanisms.
This paper introduces VectraYX-Vision-1B, a sub-2B Spanish/LATAM cybersecurity vision-language model, but reports a negative visual-grounding result, raising architectural questions about NoPE layers and releasing code, benchmarks, and checkpoints.
This paper studies PEFT-BD, a speculative decoding method using a LoRA-like adapter as a block-diffusion drafter, and finds that despite nontrivial accepted prefixes, it does not yield speedup because the drafter still requires a full-backbone pass, making it not compute-efficient.
This paper investigates whether per-phase metric selection improves demonstration curation for behavior cloning policies. The authors find that phase-gated curation never outperforms global or uniform metric application, and the dilution of defect signals across phases explains the failure.
This paper investigates whether direct activation transfer between language models can improve reasoning, using a linear translation layer from Pythia-160M to Pythia-410M. Despite achieving high representational alignment, the transferred activations do not improve multi-hop question answering, yielding a negative result.