The central challenge of reasoning-intensive retrieval lies in identifying implicitreasoning relationships between queries and documents, rather than superficial se-mantic or lexical similarity. The contrastive learning paradigm is fundamentallya static representation consolidation technique: during training, it encodes hier-archical relevance concepts into fixed geometric structures in the vector space,and at inference time it cannot dynamically adjust relevance judgments accord-ing to the specific reasoning demands of each query. Consequently, performancedegrades noticeably when vocabulary mismatch exists between queries and doc-uments or when implicit reasoning is required to establish relevance. This pa-per proposes Thought 1 (T1), a generative retrieval model that shifts relevancemodeling from static alignment to dynamic reasoning. On the query side, T1 dy-namically generates intermediate reasoning trajectories for each query to bridgeimplicit reasoning relationships and uses <embtoken> as a semantic aggregationpoint for the reasoning output. On the document side, it employs an instruction+ text + <embtoken> encoding format to support high-throughput indexing. Tointernalize dynamic reasoning capabilities into vector representations, we adopt athree-stage training curriculum and introduce GRPO in the third stage, enablingthe model to learn optimal derivation strategies for different queries through trial-and-error reinforcement learning. On the BRIGHT benchmark, T1-4B exhibitsstrong performance under the original query setting, outperforming larger modelstrained with contrastive learning overall, and achieving performance comparableto multi-stage retrieval pipelines. The results demonstrate that replacing static rep-resentation alignment with dynamic reasoning generation can effectively improvereasoning-intensive retrieval performance.
Figures & tables
Figure 1: Overview of the T1 model. Top: paradigm comparison between contrastive learning (static) and T1 (dynamic); Bottom: three-stage curriculum training pipeline (see Section 3.2 for details).
Models
StackExchange
Code
Theorem-based
Avg
Bio
Earth
Econ
Psy
Rob
Stack
Sus
Pony
Leet
AoPS
TheoQ
TheoT
Evaluate Retriever with Original Query
ReasonIR-8B
26.2
31.4
23.3
30.0
18.0
23.9
20.5
10.5
35.0
14.7
31.9
27.2
24.4
RaDeR-7B
34.6
38.9
22.1
33.0
14.8
22.5
23.7
5.0
37.3
10.2
28.4
35.1
25.5
Seed1.5-Embedding
34.8
46.9
23.4
31.6
19.1
25.4
21.0
4.9
43.2
12.2
33.3
30.5
27.2
T1-4B
57.4
54.8
30.6
48.2
33.1
36.4
35.6
31.9
14.9
11.9
41.6
48.5
37.1
Table 1: Performance comparison of different models on BRIGHT
Models
StackExchange
Code
Theorem-based
Avg
Bio
Earth
Econ
Psy
Rob
Stack
Sus
Pony
Leet
AoPS
TheoQ
TheoT
T1-4B (Stage 1)
23.8
39.2
18.4
30.0
21.3
23.5
19.8
33.2
6.7
12.1
27.5
20.5
23.0
T1-4B (Stage 2)
53.8
53.6
29.5
44.5
31.8
34.5
34.8
36.6
12.7
11.1
40.7
45.1
35.7
T1-4B (Stage 3)
57.4
54.8
30.6
48.2
33.1
36.4
35.6
31.9
14.9
11.9
41.6
48.5
37.1
Table 2: Performance comparison of T1-4B across training stages on BRIGHT
Models
StackExchange
Code
Theorem-based
Avg
Bio
Earth
Econ
Psy
Rob
Stack
Sus
Pony
Leet
AoPS
TheoQ
TheoT
Contrastive Learning (4B)
48.1
51.3
31.2
44.3
27.1
30.9
34.5
24.3
14.5
8.7
37.6
46.4
33.2
T1-4B
57.4
54.8
30.6
48.2
33.1
36.4
35.6
31.9
14.9
11.9
41.6
48.5
37.1
Table 3: Performance comparison of contrastive learning and T1 on BRIGHT
Reasoning-intensive retrieval requires text representations to capture not only semantic similarity, but also the reasoning needed to determine relevance under a given retrieval instruction. Existing reasoning-enhanced embedding models improve retrieval by incorporating reasoning information into dense representations, yet their supervision is typically dominated by the final retrieval objective. As a result, latent reasoning trajectories may learn shortcut reasoning patterns that preserve retrieval performance without producing meaningful incremental retrieval gains. We propose Retrieval Grounding Latent Reasoning (RGLT), a latent reasoning framework for dense retrieval that explicitly connects intermediate latent transitions with retrieval improvements. RGLT performs non-autoregressive reasoning in hidden space through an instruction-conditioned latent reasoning trajectory constructed from silent tokens. It combines process-supervised explicit-to-implicit distillation with retrieval-grounded supervision, using stage-wise CoT reconstruction to shape intermediate latent states and retrieval-effect credit to optimize incremental retrieval gains across the latent reasoning trajectories. Experiments on reasoning-intensive retrieval benchmarks show that RGLT consistently outperforms strong baselines while preserving efficient embedding inference.
Gang Zhou, Xiongxi Yu, Hu Tian +5
School of Artificial Intelligence, Beijing University of Posts and Telecommunications · Meituan LongCat Interaction Team · School of Management Science and Engineering, Central University of Finance and Economics +1
Large reasoning models such as DeepSeek-R1 and OpenAI o1 generate extended chains of thought spanning thousands of tokens, yet their integration with retrieval-augmented generation (RAG) remains fundamentally misaligned. Current RAG systems optimize for providing context before reasoning begins, while reasoning models require evidence injection during multi-step inference chains. We introduce ReaLM-Retrieve, a reasoning-aware retrieval framework that addresses this mismatch through three key innovations: (1) a step-level uncertainty detector that identifies knowledge gaps at reasoning-step granularity rather than token or sentence level; (2) a retrieval intervention policy that learns when external evidence maximally benefits ongoing reasoning; and (3) an efficiency-optimized integration mechanism that reduces per-retrieval overhead by 3.2x compared to naive integration. Experiments on MuSiQue, HotpotQA, and 2WikiMultiHopQA demonstrate that ReaLM-Retrieve achieves on average 10.1% absolute improvement in answer F1 over standard RAG (range: 9.0-11.8% across the three benchmarks) while reducing retrieval calls by 47% compared to fixed-interval approaches like IRCoT (all improvements significant at p<0.01, paired bootstrap). On the challenging MuSiQue benchmark requiring 2-4 hop reasoning, our method achieves 71.2% F1 with an average of only 1.8 retrieval calls per question. Analysis shows that ReaLM-Retrieve also improves retrieval quality itself, achieving 81.3% Recall@5 with consistently higher precision and MRR than fixed-interval baselines on supporting evidence, establishing new state-of-the-art efficiency-accuracy trade-offs for reasoning-intensive retrieval tasks.
Dongxin Guo, Jikun Wu, Siu Ming Yiu
The University of Hong Kong Hong Kong, China · Brain Investing Limited Hong Kong, China · Stellaris AI Limited Hong Kong, China
Reasoning-conditioned embeddings can improve retrieval by progressively refining semantic constraints. However, additional reasoning may introduce over-generalized or semantically noisy abstractions that weaken retrieval specificity. Existing methods largely supervise reasoning through final retrieval outcomes, providing limited guidance over the refinement process itself. We propose AURA, a framework that directly supervises intermediate reasoning states according to how well they separate relevant documents from hard distractors. AURA further trains a lightweight utility estimator over prefix-conditioned embeddings to adapt reasoning depth at inference time. On the reasoning-intensive BRIGHT benchmark, AURA-4B achieves 34.7 average nDCG@10 while using 2.8 reasoning steps and 96 generated tokens per query. Under matched deployment conditions, adaptive stopping reduces latency from 4.6 to 0.76 seconds relative to full-trajectory reasoning. Results on MS MARCO further suggest that this specialization need not substantially degrade conventional in-domain retrieval.
Yingtao Luo, Kangyi Zhao, Shiyi Du +1
Carnegie Mellon University · University of Pittsburgh