SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2505.20674 · ICLR · 2025

PonderLM: Pretraining Language Models to Ponder in Continuous Space

Siyuan Huang, Yixuan Wang, Xinbing Wang, Zhouhan Lin, Boyi Zeng, Zhiyu Li, He Li, Ziwei He, Shixiang Song

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 6 functions out of this paper's own repositories and ran 6 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
lumia-group/ponderinglm canonical 3 of 3
copy not recorded — 3 of 3
FunctionStatusWhere it lives
repeat_kv Ran this paper's copy was not recorded; identical code first harvested from fe1ixxu/ALMA
pointer only · get_code("30d7eec482ebf6b1")
apply_rotary_pos_emb Ran this paper's copy was not recorded; identical code first harvested from fe1ixxu/ALMA
pointer only · get_code("bac65c3dafaec040")
create_score_evaluation_response Ran lumia-group/ponderinglm/src/llamafactory/api/chat.py
code served (permissive licence) · get_code("dbea5358be6b5ad0")
dictify Ran lumia-group/ponderinglm/src/llamafactory/api/common.py
code served (permissive licence) · get_code("52a9ebcf2e9dee87")
jsonify Ran lumia-group/ponderinglm/src/llamafactory/api/common.py
code served (permissive licence) · get_code("57583cf880d95444")
rotate_half Ran this paper's copy was not recorded; identical code first harvested from fe1ixxu/ALMA
pointer only · get_code("b99eea6376d1e212")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Humans ponder before articulating complex sentence elements, enabling deeper cognitive processing through focused effort. In this work, we introduce this pondering process into language models by repeatedly invoking the forward process within a single token generation step. During pondering, instead of generating an actual token sampled from the prediction distribution, the model ponders by yielding a weighted sum of all token embeddings according to the predicted token distribution. The generated embedding is then fed back as input for another forward pass. We show that the model can learn to ponder in this way through self-supervised learning, without any human annotations. Experiments across three widely used open-source architectures-GPT-2, Pythia, and LLaMA-and extensive downstream task evaluations demonstrate the effectiveness and generality of our method. On 9 downstream benchmarks, our pondering-enhanced Pythia models significantly outperform the official Pythia models. Notably, our PonderPythia models demonstrate remarkable effectiveness: PonderPythia-

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2505.20674")
get_code_for_paper("2505.20674")
have("2505.20674")

Connect an agent — have() is free.