REVIEW 3 major objections 4 minor 60 references
Relevant but Incomplete: Referential Dangling as a Paradigm-Level Failure Mode in Hard Prompt Compression
T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read Hard prompt compression can keep a relevant answer while deleting the definition that makes it interpretable; the paper shows this 'referential dangling' is systematic and that fixed-budget reselection recovers most of the accuracy.
desk verdict A real and well-documented failure mode in hard prompt compression, though the headline prevalence numbers rest on a lightly validated proxy and the automatic restoration does not yet beat random insertion. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the formal condition for referential dangling, built on inclusion-minimal sufficient support sets in the spirit of extractive summarization. For an extractive compressed output, $\operatorname{Ret}_C(A)$ is the set of source sentences fully retained; $T_{C,Q}$ is the set of task-relevant sentences; and $\mathcal{D}_{C,Q}(s)$ is the family of inclusion-minimal subsets of other sentences that, together with $s$ and the question, contain all explicit information needed to interpret and use $s$. Equation (1) says the output dangles when some retained task-relevant sentence has no fully retained sufficient support set. The matching machinery is the additive selection model: each fragment gets a scalar utility, and the selector maximizes summed utility under a token budget with no joint-retention term; Proposition 1 shows this guarantee is absent even when a complete alternative fits the budget. The empirical proxy used to apply the condition to real outputs is directional content-word overlap $\operatorname{cov}(v,A) \ge 0.5$ between an output unit and a source unit.
What would settle it
Take a random sample of several hundred flagged bridge examples, have independent annotators apply the paper's formal sufficiency condition (Equation 1) to decide whether the omitted paragraph is genuinely required to interpret the retained answer, and recompute the rates; if the true rate is near zero, or if a fresh fixed-budget reselection on new bridge data recovers well under 88% of the gap to full support, the central claim fails.
Extended reading notes
Core claim
The paper's central claim is that independent scoring cannot guarantee that a compressed prompt remains self-contained: the additive selection objective has no interaction term enforcing joint retention of a fragment and the text that interprets it. Formally, a compressed context exhibits referential dangling when it retains a task-relevant sentence $s$ but retains no member of $\mathcal{D}_{C,Q}(s)$, the family of inclusion-minimal sufficient support sets for $s$ (Equation 1). Proposition 1 shows that for three equal-cost fragments $\{x,d,z\}$ and budget $2m$, the unique additive maximizer can be $\{x,z\}$, whose output dangles, even though $\{x,d\}$ is feasible and complete. The paper then measures the phenomenon across implementations: at ratio 0.30, Beaver dangles 34.2-54.2% of bridge examples across three datasets; six compressors score 32.1-59.8% on a shared HotpotQA bridge set; and every LongBench-v2 document has at least one dangling reference. Fixed-budget reselection on dangling examples improves Qwen3-8B accuracy by 28.8-34.3 points and recovers at least 88% of the gap to full support, while GPT-5.5 remains 8.8 points less accurate on compressed MuSiQue contexts.
Load-bearing premise
The load-bearing premise is that 50% content-word overlap between a retained unit and the original supporting text really means the support is present; if that proxy is too lenient or too strict, the dangling rates and the every-document LongBench-v2 result change.
Editorial extensions
If this is right
- At compression ratio 0.30, a query-aware chunk compressor leaves the answer path incomplete in 34-54% of bridge examples across HotpotQA, 2WikiMultiHopQA, and MuSiQue.
- Across six hard compressors using different scoring signals, the dangling rate on a shared HotpotQA bridge set ranges from 32.1% to 59.8%, so the failure is not tied to one selection method.
- Every one of the 80 LongBench-v2 Single-Document QA documents contains at least one dangling reference under the first-mention diagnostic, with rates from 25% to 37% across seven subdomains.
- Reinserting the missing supporting paragraph while removing nonsupporting paragraphs at the same token budget improves Qwen3-8B accuracy by 28.8-34.3 points on dangling examples and recovers at least 88% of the gap to retaining both supporting paragraphs.
- Automatic restoration with a small classifier adds 4.7 points on HotpotQA with Qwen3-8B while the compression ratio changes only from 0.30 to 0.31.
Reading between the lines
- The same split-evidence failure should appear in retrieval-augmented generation, where scored or truncated passages can keep an answer string while dropping the passage that defines a referenced entity; the LongBench-v2 first-mention result is early evidence of that wider reach.
- A direct design implication, not built in the paper, is to add a joint-retention bonus to the selection objective—score a fragment together with the text that interprets it—instead of restoring omitted sentences as a separate pass.
- The automatic-restoration gain is probably limited by candidate construction rather than by the classifier, because a diagnostic condition that put the annotated supporting sentence in the candidate set recovered 8.0 points versus 4.7 for first-mention candidates.
- Reported prevalence should be read as proxy measurements: a human-annotated application of the formal condition could shift the exact rates, especially the every-document LongBench-v2 result.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper identifies a failure mode of hard prompt compression: independent scoring can retain a relevant fragment while deleting a dependency needed to interpret it, which it calls referential dangling. It formalizes the notion in Section 3, proves an existence result (Proposition 1), and then measures prevalence using a directional content-word overlap proxy. The reported empirical claims are that Beaver at compression ratio 0.30 leaves the answer path incomplete in 34-54% of bridge examples across HotpotQA, 2WikiMultiHopQA, and MuSiQue; that all six hard compressors tested exhibit dangling rates from 32% to 60% on a shared HotpotQA bridge set; that every LongBench-v2 Single-Document QA document contains at least one dangling reference; that fixed-budget reselection improves Qwen3-8B accuracy by 29-34 points on dangling examples; and that a trained classifier can restore some of the loss automatically, improving accuracy by 4.7 points on HotpotQA while changing the compression ratio from 0.30 to 0.31. The paper concludes that hard compressors should optimize both relevance and referential completeness.
Significance. If the prevalence estimates hold, this is an important and actionable observation: it identifies a structural failure shared across scoring signals, and the fixed-budget accuracy gains suggest that the phenomenon has practical cost. The paper has genuine strengths: a clear formal definition with a self-contained existence proof, a controlled fixed-budget comparison with exact McNemar tests, robustness sweeps over the overlap threshold and compression ratio, exact model identifiers in Table 9, and released code, results, and trained models. However, the central prevalence claims rest on a proxy that is only partially validated, and the main intervention changes multiple aspects of the context simultaneously. These issues are load-bearing because the paper's headline contribution is that referential dangling is a systematic, paradigm-level failure rather than a rare artifact.
major comments (3)
- [Section 4, Measurement protocol] The headline prevalence rates in Table 1, Figure 3, and the LongBench-v2 result are not direct measurements of the sufficiency condition in Eq. (1). They rely on directional content-word overlap with threshold 0.5 and, for the bridge datasets, check only the paired definition paragraph without enumerating alternative support paths that Eq. (1) explicitly allows. The manuscript acknowledges both limitations ('This protocol approximates Equation (1) rather than evaluating it exactly' and 'This test does not enumerate alternative support paths'). The 40-example human audit in Appendix B estimates precision on flagged cases only (38/40); it does not measure false negatives among unflagged examples, and the Appendix A threshold sweep shows stability of the proxy, not that the proxy tracks exact referential completeness. Because the central claim that dangling is systematic is a prevalence claim, please add a random-sample audit that applies exact sentence retention or human sufficiency judgments to both flagged and unflagged examples and reports precision and recall, or explicitly downgrade the prevalence claims to proxy-based estimates.
- [Section 4.1, Long documents] The first-mention diagnostic counts any retained later reference whose first-mention sentence is absent, regardless of whether that reference is needed to answer the query. This does not instantiate Eq. (1), which requires the dangling sentence to be in T_{C,Q}. Consequently the statements 'every document in LongBench-v2 Single-Document QA contains at least one dangling reference' and the mean rate of 30.5% are not evidence for the paper's query-dependent notion of referential dangling. Please restrict the diagnostic to question-relevant entities or references, or relabel these numbers as 'omitted first mentions' rather than 'dangling references' in the abstract and conclusion.
- [Section 5, Comparison protocol] Panel 1 of Table 3 shows large gains for Reselected over Base, but the intervention changes two things at once: it reinserts the missing supporting paragraph and removes the lowest-scoring nonsupporting paragraphs. The paper acknowledges this ('these comparisons measure the joint change in retained content rather than the isolated contribution of the reinserted paragraph'). Without an ablation that removes nonsupporting paragraphs without reinsertion, or reinserts with a budget increase, the experiment does not establish that referential completeness, rather than distractor removal, drives the 28.8-34.3 point gains. The matched addition control in Appendix G (+2.0 points for random insertion versus +4.7 for targeted restoration) is suggestive but the direct contrast is not significant (p=0.15). Please add the missing ablation or weaken the causal conclusion about the role of completing the missing support.
minor comments (4)
- [Abstract and Section 4.1] The abstract says Beaver leaves the answer path incomplete in 34-54% of bridge examples, but Table 1's rho_d is conditional on the answer paragraph satisfying Keep_0.5; please state the denominator explicitly as 'bridge examples with a retained answer paragraph' to avoid overstating the proportion of all bridge examples.
- [Section 6 and Appendix F] The main K=3 result was selected after sweeping K, and the reported p=0.022 does not account for this selection. Please report the K sweep with a multiple-comparison caveat or state that K=3 was chosen by a held-out criterion.
- [Appendix D] The McNemar tests are described as exact and two-sided in Appendix D; please state this in the main text near Table 3, since the p-values are central to the fixed-budget claim.
- [Appendix E] The DAC proxy row in Table 9 lists Qwen3-0.6B as both the scoring and proxy model; a brief sentence explaining the role of the proxy would improve reproducibility.
Circularity Check
No circularity: Proposition 1 is a self-contained existence proof, and the prevalence, reselection, and restoration claims are measured against external datasets and compressor outputs rather than derived from the paper's own definitions or fitted parameters.
full rationale
The paper's derivation chain is self-contained at every load-bearing step. Referential dangling is defined formally in Eq. (1) with stated assumptions (explicit-only support, D_{C,Q}(s) nonempty), and Proposition 1 is an existence proof whose construction parameters (u(x) > u(z) > u(d) >= 0, tau = 2m) are fully stated and whose proof does not draw on any empirical result of the paper. The prevalence claims (34-54% for Beaver, 32-60% across six compressors, 'every document' on LongBench-v2) are measurements made with a directional content-word-overlap proxy that the paper explicitly calls an approximation ('This protocol approximates Equation (1) rather than evaluating it exactly'); the proxy is not fitted to produce the headline rates (at r=0.70 the same diagnostic gives 6.3%), is swept over thresholds 0.3-0.7 in Appendix A, and is audited on 40 flagged examples with 95% precision. The fixed-budget reselection experiment is an oracle intervention driven by external supporting-paragraph annotations; the 29-34 point accuracy improvement is measured rather than derived, and the paper verifies that reselected contexts are slightly shorter, so the gain is not a token-budget artifact. The restoration classifier is trained on the HotpotQA training split, evaluated on disjoint contexts, and honestly controlled (random insertion gives +2.0 vs. +4.7 targeted, p=0.15, which the paper states 'does not establish an advantage over random insertion'). The one centrally placed author self-citation is Beaver (Hu et al. 2026), the system under evaluation, which is exercised via its official released implementation and is therefore code-reproduced and externally falsifiable; it is the object of study, not the justification for the claims. Genuine concerns remain as measurement-validity issues, not circularity: the LongBench-v2 first-mention diagnostic counts any omitted first-mention regardless of question-relevance (a looseness the paper acknowledges), and the K=3 restoration setting is the largest observed gain on the evaluation set (a multiple-comparisons concern). Neither concern reduces a claimed result to its inputs by construction.
Assumptions & free parameters
free parameters (2)
- content-word overlap threshold theta =
0.5
- restoration count K =
3
assumptions (3)
- domain assumption For every task-relevant sentence s in T_C,Q, a sufficient support set D_C,Q(s) is nonempty
- ad hoc to paper Directional content-word overlap with threshold 0.5 approximates the exact sufficiency condition of Equation (1)
- domain assumption Extractive hard compressors with nonoverlapping units are modeled by the additive knapsack objective (Eq. 4)
Cite this review
Pith. "Pith review of Relevant but Incomplete: Referential Dangling as a Paradigm-Level Failure Mode in Hard Prompt Compression." pith.science (2026). https://pith.science/paper/73L2HPLP
@misc{pith2026260804569,
author = {Pith},
title = {Pith review of: Relevant but Incomplete: Referential Dangling as a Paradigm-Level Failure Mode in Hard Prompt Compression},
year = {2026},
howpublished = {\url{https://pith.science/paper/73L2HPLP}},
note = {Machine review of arXiv:2608.04569}
}
read the original abstract
Hard prompt compression reduces long-context inference cost by independently scoring tokens, sentences, or chunks and retaining the highest-scoring units under a budget. We identify a structural failure in this procedure: independent selection can split dependent evidence pairs, retaining one member while deleting the other. When retained text contains an answer but deleted text defines the entity needed to interpret it, we call the result referential dangling. At a compression ratio of 0.30, Beaver, which ranks coherent chunks using Qwen3-0.6B embeddings, leaves the answer path incomplete in 34-54% of bridge examples across three multi-hop question answering datasets. On a shared HotpotQA bridge set, all six hard compressors we test exhibit dangling at rates up to 60%, and every document in LongBench-v2 Single-Document QA contains at least one dangling reference. On dangling examples evaluated with Qwen3-8B, reinserting the missing supporting paragraph while removing nonsupporting paragraphs to maintain the token budget improves accuracy by 29-34 percentage points (p < 0.0001), recovering at least 88% of the gap to contexts retaining both supporting paragraphs. Stronger answer models do not absorb the loss: on MuSiQue, GPT-5.5 is 8.8 points less accurate on compressed contexts than on contexts retaining both supporting paragraphs. Finally, we train a compact classifier to rank omitted sentences by whether they are needed to interpret retained text and reinsert the top-ranked candidates without support annotations at inference. On HotpotQA with Qwen3-8B, this automatic restoration improves accuracy by 4.7 points while changing the compression ratio only from 0.30 to 0.31. Hard compressors should optimize both relevance and referential completeness.
Figures
Figures from the paper (3 more)
Reference graph
Works this paper leans on
-
[1]
Transactions of the association for computational linguistics , volume=
Lost in the middle: How language models use long contexts , author=. Transactions of the association for computational linguistics , volume=
-
[2]
International Conference on Machine Learning , pages=
Large language models can be easily distracted by irrelevant context , author=. International Conference on Machine Learning , pages=. 2023 , organization=
2023
-
[3]
Advances in neural information processing systems , volume=
Flashattention: Fast and memory-efficient exact attention with io-awareness , author=. Advances in neural information processing systems , volume=
-
[4]
Proceedings of the 2023 conference on empirical methods in natural language processing , pages=
Llmlingua: Compressing prompts for accelerated inference of large language models , author=. Proceedings of the 2023 conference on empirical methods in natural language processing , pages=
2023
-
[5]
Findings of the Association for Computational Linguistics: ACL 2024 , pages=
Llmlingua-2: Data distillation for efficient and faithful task-agnostic prompt compression , author=. Findings of the Association for Computational Linguistics: ACL 2024 , pages=
work page 2024
-
[6]
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
Longllmlingua: Accelerating and enhancing llms in long context scenarios via prompt compression , author=. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
-
[7]
Proceedings of the 2023 conference on empirical methods in natural language processing , pages=
Compressing context to enhance inference efficiency of large language models , author=. Proceedings of the 2023 conference on empirical methods in natural language processing , pages=
2023
-
[8]
arXiv preprint arXiv:2603.19635 , year=
BEAVER: A Training-Free Hierarchical Prompt Compression Method via Structure-Aware Page Selection , author=. arXiv preprint arXiv:2603.19635 , year=
Show all 60 references
-
[9]
IEEE Transactions on Pattern Analysis and Machine Intelligence , year=
Parse trees guided LLM prompt compression , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=
-
[10]
Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
DAC: A dynamic attention-aware approach for task-agnostic prompt compression , author=. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
-
[11]
International Conference on Learning Representations , volume=
Recomp: Improving retrieval-augmented lms with context compression and selective augmentation , author=. International Conference on Learning Representations , volume=
-
[12]
IEEE Access , volume=
Discrete prompt compression with reinforcement learning , author=. IEEE Access , volume=. 2024 , publisher=
2024
-
[13]
Proceedings of the AAAI Conference on Artificial Intelligence , volume=
Prompt compression with context-aware sentence encoding for fast and improved llm inference , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=
-
[14]
Advances in Neural Information Processing Systems , volume=
Learning to compress prompts with gist tokens , author=. Advances in Neural Information Processing Systems , volume=
-
[15]
Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages=
Adapting language models to compress contexts , author=. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages=
2023
-
[16]
arXiv preprint arXiv:2307.06945 , year=
In-context autoencoder for context compression in a large language model , author=. arXiv preprint arXiv:2307.06945 , year=
-
[17]
Prompt compression for large language models: A survey , author=. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers) , pages=
2025
-
[18]
ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=
Subnetwork-to-go: Elastic neural network with dynamic training and customizable inference , author=. ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=. 2024 , organization=
2024
-
[19]
arXiv preprint arXiv:2209.15200 , year=
An efficient encoder-decoder architecture with top-down attention for speech separation , author=. arXiv preprint arXiv:2209.15200 , year=
-
[20]
2025 IEEE International Conference on Multimedia and Expo (ICME) , pages=
SPMamba: Leveraging Long-Sequence Modeling with State Space Models for Speech Separation , author=. 2025 IEEE International Conference on Multimedia and Expo (ICME) , pages=. 2025 , organization=
2025
-
[21]
arXiv preprint arXiv:2509.23610 , year=
Efficient Audio-Visual Speech Separation with Discrete Lip Semantics and Multi-Scale Global-Local Attention , author=. arXiv preprint arXiv:2509.23610 , year=
-
[22]
Proceedings of the 2018 conference on empirical methods in natural language processing , pages=
HotpotQA: A dataset for diverse, explainable multi-hop question answering , author=. Proceedings of the 2018 conference on empirical methods in natural language processing , pages=
2018
-
[23]
Proceedings of the 28th International Conference on Computational Linguistics , pages=
Constructing a multi-hop qa dataset for comprehensive evaluation of reasoning steps , author=. Proceedings of the 28th International Conference on Computational Linguistics , pages=
-
[24]
Transactions of the Association for Computational Linguistics , volume=
MuSiQue: Multihop Questions via Single-hop Question Composition , author=. Transactions of the Association for Computational Linguistics , volume=. 2022 , publisher=
2022
-
[25]
Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
Longbench v2: Towards deeper understanding and reasoning on realistic long-context multitasks , author=. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
-
[26]
arXiv preprint arXiv:2404.06654 , year=
RULER: What's the real context size of your long-context language models? , author=. arXiv preprint arXiv:2404.06654 , year=
-
[27]
Advances in Neural Information Processing Systems , volume=
H2o: Heavy-hitter oracle for efficient generative inference of large language models , author=. Advances in Neural Information Processing Systems , volume=
-
[28]
International Conference on Learning Representations , volume=
Efficient streaming language models with attention sinks , author=. International Conference on Learning Representations , volume=
-
[29]
Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
Learning-based single-document summarization with compression and anaphoricity constraints , author=. Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
-
[30]
Proceedings of the 2017 conference on empirical methods in natural language processing , pages=
End-to-end neural coreference resolution , author=. Proceedings of the 2017 conference on empirical methods in natural language processing , pages=
2017
-
[31]
Proceedings of the 58th annual meeting of the association for computational linguistics: system demonstrations , pages=
Stanza: A Python natural language processing toolkit for many human languages , author=. Proceedings of the 58th annual meeting of the association for computational linguistics: system demonstrations , pages=
-
[32]
Proceedings of the 58th annual meeting of the association for computational linguistics , pages=
On faithfulness and factuality in abstractive summarization , author=. Proceedings of the 58th annual meeting of the association for computational linguistics , pages=
-
[33]
Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP) , pages=
Dense passage retrieval for open-domain question answering , author=. Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP) , pages=
2020
-
[34]
Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval , pages=
Colbert: Efficient and effective passage search via contextualized late interaction over bert , author=. Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval , pages=
-
[35]
Advances in neural information processing systems , volume=
Retrieval-augmented generation for knowledge-intensive nlp tasks , author=. Advances in neural information processing systems , volume=
-
[36]
arXiv preprint arXiv:2407.21783 , year=
The llama 3 herd of models , author=. arXiv preprint arXiv:2407.21783 , year=
-
[37]
arXiv preprint arXiv:2505.09388 , year=
Qwen3 technical report , author=. arXiv preprint arXiv:2505.09388 , year=
-
[38]
2023 , eprint=
Mistral 7B , author=. 2023 , eprint=
2023
-
[39]
Bert: Pre-training of deep bidirectional transformers for language understanding , author=. Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers) , pages=
2019
-
[40]
arXiv preprint arXiv:2506.05176 , year=
Qwen3 embedding: Advancing text embedding and reranking through foundation models , author=. arXiv preprint arXiv:2506.05176 , year=
-
[41]
arXiv preprint arXiv:2307.09288 , year=
Llama 2: Open foundation and fine-tuned chat models , author=. arXiv preprint arXiv:2307.09288 , year=
-
[42]
Findings of the association for computational linguistics: EMNLP 2023 , pages=
Tcra-llm: Token compression retrieval augmented large language model for inference cost reduction , author=. Findings of the association for computational linguistics: EMNLP 2023 , pages=
2023
-
[43]
Findings of the Association for Computational Linguistics: ACL 2024 , pages=
Extending context window of large language models via semantic compression , author=. Findings of the Association for Computational Linguistics: ACL 2024 , pages=
2024
-
[44]
Findings of the Association for Computational Linguistics: ACL 2025 , pages=
Taco-rl: Task aware prompt compression optimization with reinforcement learning , author=. Findings of the Association for Computational Linguistics: ACL 2025 , pages=
2025
-
[45]
Expert Systems with Applications , volume=
Prompt Compression based on Key-Information Density , author=. Expert Systems with Applications , volume=. 2025 , publisher=
2025
-
[46]
arXiv preprint arXiv:2407.08892 , year=
Characterizing prompt compression methods for long context inference , author=. arXiv preprint arXiv:2407.08892 , year=
-
[47]
Advances in Neural Information Processing Systems , volume=
Fundamental limits of prompt compression: A rate-distortion framework for black-box language models , author=. Advances in Neural Information Processing Systems , volume=
-
[48]
International Conference on Learning Representations , volume=
Long context compression with activation beacon , author=. International Conference on Learning Representations , volume=
-
[49]
Findings of the Association for Computational Linguistics: EMNLP 2024 , pages=
From reading to compressing: Exploring the multi-document reader for prompt compression , author=. Findings of the Association for Computational Linguistics: EMNLP 2024 , pages=
2024
-
[50]
Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
A silver bullet or a compromise for full attention? a comprehensive study of gist token-based context compression , author=. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
-
[51]
ICASSP 2026-2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=
Dspc: Dual-stage progressive compression framework for efficient long-context reasoning , author=. ICASSP 2026-2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages=. 2026 , organization=
2026
-
[52]
European Conference on Information Retrieval , pages=
Prompt Compression in the Wild: Measuring Latency, Rate Adherence, and Quality for Faster LLM Inference , author=. European Conference on Information Retrieval , pages=. 2026 , organization=
2026
-
[53]
Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
500xcompressor: Generalized prompt compression for large language models , author=. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=
-
[54]
arXiv preprint arXiv:2503.19114 , year=
Understanding and improving information preservation in prompt compression for llms , author=. arXiv preprint arXiv:2503.19114 , year=
-
[55]
Findings of the Association for Computational Linguistics: NAACL 2025 , pages=
Perception compressor: A training-free prompt compression framework in long context scenarios , author=. Findings of the Association for Computational Linguistics: NAACL 2025 , pages=
2025
-
[56]
Advances in Neural Information Processing Systems , volume=
Efficient prompt compression with evaluator heads for long-context transformer inference , author=. Advances in Neural Information Processing Systems , volume=
-
[57]
Information Processing & Management , volume=
SelfCP: Compressing over-limit prompt via the frozen large language model itself , author=. Information Processing & Management , volume=. 2024 , publisher=
2024
-
[58]
arXiv preprint arXiv:2512.14244 , year=
From Context to EDUs: Faithful and Structured Context Compression via Elementary Discourse Unit Decomposition , author=. arXiv preprint arXiv:2512.14244 , year=
-
[59]
OpenAI blog , volume=
Language models are unsupervised multitask learners , author=. OpenAI blog , volume=
-
[60]
2026 , howpublished =
2026
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.