NLP & Language
Research on NLP & Language: explanations, evidence, limitations and connections between papers.
12 results
Judge disagreement clarifies rubrics but not human alignment
Disagreement-guided rubric edits raised judge agreement on news summaries, with no detectable gain in expert match.
A new user turn helps redirect failed reasoning
UPTURN redirects failed reasoning from a learned point and improves average accuracy across four reasoners on six math tests.
POrchestra lets agents revise task setups mid-task
On simulated app tasks with updates, POrchestra had higher valid-scenario success than ReAct under the same execution budget.
Parallel draft correction raises decoding speed without sequential repair
A contextual corrector improves speculative-decoding throughput while keeping candidate scoring parallel in tested serving settings.
A crisis-support test shows recognition is not enough
A clinician-authored crisis-support test finds models often recognize danger without connecting users to human help.
One language model gains accuracy by spending more inference loops
Elastic Latent Recurrence keeps one-loop scores near matched baselines and raises ten-task averages with additional loops.
Choosing fewer, better tokens improves distillation
DIAL-OPD ranks response tokens by disagreement and probability, improving mean accuracy over full-token distillation in math tests.
Byte models miss results that need calculation
A byte-level model’s loss-based patch signal finds computed results that entropy misses on arithmetic text.
A cache compressor rehearses likely answers before questions arrive
Attic-KV rehearses likely quoted answers and context anchors to improve cache compression before future questions are known.
TokenRouter helps models hand off requests without restarting work
TokenRouter coordinates model handoffs and raised output-token rates over existing systems in tested workloads.
Reusing a student's first answers can improve distillation
Semi-on-policy distillation reuses a student's starting answers and outperformed standard distillation on 14 of 17 tested pairs.
An optimizer change slows forgetting during adaptation
A training-time optimizer adjustment reduced old-domain validation-loss increases in seven stable Korean adaptation settings.