LaTER: Efficient Test-Time Reasoning via Latent Exploration and Explicit Verification

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Xuan, Wang, Yining, Liu, Yuchen, Liu, Guanjun, Qiu, Delai, Liu, Shengping, Liang, Jiaen, Huang, Wei, Yu, Jun, Zhu, Junnan
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910200670191616
author Li, Xuan
Wang, Yining
Liu, Yuchen
Liu, Guanjun
Qiu, Delai
Liu, Shengping
Liang, Jiaen
Huang, Wei
Yu, Jun
Zhu, Junnan
author_facet Li, Xuan
Wang, Yining
Liu, Yuchen
Liu, Guanjun
Qiu, Delai
Liu, Shengping
Liang, Jiaen
Huang, Wei
Yu, Jun
Zhu, Junnan
contents Chain-of-thought (CoT) reasoning improves large language models (LLMs) on difficult tasks, but it also makes inference expensive because every intermediate step must be generated as a discrete token. Latent reasoning reduces visible token generation by propagating continuous states, yet replacing explicit derivations with latent computation can hurt tasks that require symbolic checking. We propose Latent-Then-Explicit Reasoning (LaTER), a two-stage paradigm that first performs bounded exploration in a continuous latent space and then switches to explicit CoT for verification and answer generation. In a training-free instantiation, LaTER projects final-layer hidden states back to the input embedding space, preserves the latent KV cache, and uses entropy and model-native stop-token probes to decide when to switch. We find that strong reasoning models already exhibit structured latent trajectories under this interface. On Qwen3-14B, training-free LaTER reduces total token usage by 16%-32% on several benchmarks while matching or improving accuracy on most of them; for example, it improves AIME 2025 from 70.0% to 73.3% while reducing tokens from 15,730 to 10,661. We further construct Latent-Switch-69K, a supervised corpus that pairs condensed solution intuitions with shortened explicit derivations. Fine-tuning with latent rollout and halting supervision yields additional gains: trained LaTER reaches 80.0% accuracy on AIME 2025, 10.0 points above the standard CoT baseline, while using 33% fewer tokens. Our code, data, and model are available at https://github.com/TioeAre/LaTER.
format Preprint
id arxiv_https___arxiv_org_abs_2605_07315
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle LaTER: Efficient Test-Time Reasoning via Latent Exploration and Explicit Verification
Li, Xuan
Wang, Yining
Liu, Yuchen
Liu, Guanjun
Qiu, Delai
Liu, Shengping
Liang, Jiaen
Huang, Wei
Yu, Jun
Zhu, Junnan
Computation and Language
Chain-of-thought (CoT) reasoning improves large language models (LLMs) on difficult tasks, but it also makes inference expensive because every intermediate step must be generated as a discrete token. Latent reasoning reduces visible token generation by propagating continuous states, yet replacing explicit derivations with latent computation can hurt tasks that require symbolic checking. We propose Latent-Then-Explicit Reasoning (LaTER), a two-stage paradigm that first performs bounded exploration in a continuous latent space and then switches to explicit CoT for verification and answer generation. In a training-free instantiation, LaTER projects final-layer hidden states back to the input embedding space, preserves the latent KV cache, and uses entropy and model-native stop-token probes to decide when to switch. We find that strong reasoning models already exhibit structured latent trajectories under this interface. On Qwen3-14B, training-free LaTER reduces total token usage by 16%-32% on several benchmarks while matching or improving accuracy on most of them; for example, it improves AIME 2025 from 70.0% to 73.3% while reducing tokens from 15,730 to 10,661. We further construct Latent-Switch-69K, a supervised corpus that pairs condensed solution intuitions with shortened explicit derivations. Fine-tuning with latent rollout and halting supervision yields additional gains: trained LaTER reaches 80.0% accuracy on AIME 2025, 10.0 points above the standard CoT baseline, while using 33% fewer tokens. Our code, data, and model are available at https://github.com/TioeAre/LaTER.
title LaTER: Efficient Test-Time Reasoning via Latent Exploration and Explicit Verification
topic Computation and Language
url https://arxiv.org/abs/2605.07315