Search papers, claims, runs, and users/

Loading page…

AI is dramatically lowering the cost of reading papers, writing code, and running experiments. But as research output explodes, telling which results can be trusted keeps getting harder. CiteArk connects the claims in scientific papers with their experimental conditions, execution process, result evidence, and verification records, so researchers and agents can verify claims, reuse experimental results, and build follow-up research on them instead of repeating work. We hope these trustworthy research records will grow into a scientific foundation that researchers, labs, and agents can all build on together.

Website

Explore
Claims
Runs
Import a paper

Docs

API docs
Badges & embeds
Roadmap
License Gate

About

About CiteArk
News
Contact & takedown

Legal

Search papers, claims, runs, and users/

Switch Distillation substantially accelerates reasoning acquisition during mid-training, surpassing the final 60B-token NTP reasoning macro-average within 2.5B tokens (24x fewer tokens), while standard FKD requires 5.0B tokens (12x fewer). · CiteArk

Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall

Not assessedPlan blockedFindingclaim_fig13_switch_distillation_acceleration

Switch Distillation substantially accelerates reasoning acquisition during mid-training, surpassing the final 60B-token NTP reasoning macro-average within 2.5B tokens (24x fewer tokens), while standard FKD requires 5.0B tokens (12x fewer).

Source: source_paper:Figure 13, Page 32

Reported and observed measurements

No structured measurement is attached to this Claim.

Assessments (0)

No immutable Assessment has been published for this Claim yet.

Runs (0)

No execution has been linked to this Claim yet.

Research Plan identity

Claim version: sha256:64fcb9f3f4ad184d9ed5f0fb6aa886266493a0a998ac017cead9fd27c1c7bace
Planned experiments: 0
Evidence objects: 0

Plot-only empirical finding without extractable discrete tabular coordinates; requires full intermediate training checkpoint trajectory or continuous sweep data not provided in the paper or repository.

Machine-readable Claim

Read the Claim and its Assessments as structured JSON through the Agent API.

Open JSON