NLP & Language
Research on NLP & Language: explanations, evidence, limitations and connections between papers.
36 results
Match instruction changes to the behavior they need to establish
A changed instruction can help a system complete a task, or it can reveal that a decision depends on wording. A robot’s request, a stalled math solution, and a guard reviewing a tool call make those purposes easy to confuse. The studies offer a bounded way to separate them: test steering by the task outcome it seeks to improve, and test sensitivity by changing one input feature while keeping the case fixed. The result belongs to the model, task, and comparison that produced it.
Intern-Decision scores structured choices in one pass
Intern-Decision scores structured choices directly; its 4B model has a higher mean accuracy than Jev across seven evaluated suites.
Multi-SFT builds permissive instruction data across five languages
Models trained on Multi-SFT had higher multilingual test averages than official instruction-tuned models in six of nine comparisons.
Adding context to safety examples reduces cued misalignment
ContextLift reduced cue-triggered misalignment in three advice domains on one tested language model.
A method steers answers as it revises question context
EMO combines question-specific context revision and token steering, improving rule-scored accuracy over zero-shot generation across its tested grid.
Long-context midtraining raises selected coding scores
An open midtraining recipe improved selected post-SFT coding test scores on two Nemotron models, with results varying by test.
Strong models can review agent steps instead of writing each
SpecAgent has a larger model review grounded tool drafts, matching Strong-only success on tested Airline tasks while using less estimated memory traffic.
Math overlap rules can remove many distinct problems
An evaluation of math-data cleaning finds case-by-case adjudication better balances reliability and coverage on reviewed candidate pairs.
Read the evaluation protocol before trusting a benchmark score
Imagine seeing different accuracy scores for the same video model on the same benchmark. The model weights may be unchanged, yet the evaluator can choose different frames, process them differently, or score the response by a different rule. Across video questions, Taiwan knowledge questions, and dependency repairs, these studies show why a result needs its protocol beside it. A compact test can make a comparison less costly in its tested setting, while its task coverage still limits the claim. For a repair, success also depends on which checks the verifier runs.
Match saved computation to changing inputs and measured outcomes
A sparse calculation may face a different set of changed input rows at every step. A robot may act on a scene that moves while it computes. Reusing prior work seems attractive in both cases, but the work that remains useful has a different lifetime in each system. The practical question is what a shortcut saves under the tested workload and what outcome the decision needs. Less computation can reduce measured runtime, while task success requires its own check. These studies examine those choices in specific systems; they do not establish a shared reduction in stored memory.
Dedicated model leads on Indian language speech tests
A dedicated model leads on average in shared Indian-language speech tests, with exceptions on long read speech.
A speech model confines feature matching to protect English
A multilingual speech method confines feature matching to adapters, lowering word error rates in reported English and multilingual test groups.
A tokenizer reuses chat history to cut first-token delays
Delta Tokenizer reuses session text to cut first-token delay in tested long-context SGLang chat requests.
HybridAgent combines latent reasoning with tool use in text
HybridAgent combines latent reasoning with tool interaction; staged math tests show gains, while a proof bounds cross-turn memory claims.
JetSpec++ drafts ahead while the target checks current tokens
Across seven tests, JetSpec++ averaged 8.94 committed tokens per round versus DSpark’s 6.89 on Qwen3-8B at temperature zero and budget 16.
Sparse attention stays close to full attention at long context
JSSA selects context tokens directly, staying near full-cache language-modeling scores while speeding decoding in tested setups.
A Brazilian Portuguese model adds a tested safety step
A Brazilian Portuguese model adds safety training and refuses most harmful prompts in a 218-item evaluation.
Only two agent-training methods beat GRPO on average
A comparison of ten text-agent training methods finds average gains are limited and do not consistently outweigh training cost.
OpenTWBench tests Taiwan knowledge and exposes score effects
OpenTWBench tests Taiwan knowledge and shows how answer-reading rules can sharply change measured accuracy.
Long agent tasks gain verified progress feedback
Tev-Gym adds verified progress feedback to long agent tasks, improving full solves on an 18-task computing-and-math test.
A vision-language reward model learns from caption edits
A label-free vision-language scorer learns from caption edits; verifier-trained selection reduces absent object mentions in tested captions.
Virtual Head Attention cuts cache while preserving model quality
Small-model tests found higher average accuracy; at 1.7B, Virtual Head Attention cut cache use versus GQA-8.
Solo scores often miss which models help a team
Compositional reinforcement learning improved team performance across tested tasks and also helped in selected held-out settings.
A schema draft helps route questions across private data owners
Schema-based routing was not significantly less accurate than the best fixed protocol in ten tested settings.