Skip to content
OpenTrain AIFor AI Companies

HFEPX · Eval paper review

Learning When to Attend: Conditional Memory Access for Long-Context LLMs

Sakshi Choudhary, Aditya Chattopadhyay, Luca Zancato, Elvis Nunez +3 more

Published

Mar 18, 2026

Citations

0

Trust level

Low

Usefulness score

25/100 (Low)

Extraction confidence

45% (Low)

Derived from extracted protocol signals and abstract evidence.

Rater population

Not reported

Signals refreshed

Mar 18, 2026

Open arXiv Find implementation Shortlist (0) RSS feed

Tags

Automatic Metrics General Long Horizon

Should you rely on this paper?

This paper is adjacent to HFEPX scope and is best used for background context, not as a primary protocol reference.

Use this as background context only. Do not make protocol decisions from this page alone.

Best use

Background context only

Use if you need

A secondary eval reference to pair with stronger protocol papers.

What to verify

Validate the evaluation procedure and quality controls in the full paper before operational use.

Main weakness

The available metadata is too thin to trust this as a primary source.

Human feedback signal
Not explicit
Not explicit in abstract metadata
Evaluation signal
Detected
Eval setup described
Usefulness for eval research
25/100
Adjacent candidate

Treat as adjacent context, not a core eval-method reference.

If you are doing eval pipeline work, start here

Human Eval Papers LLM-as-Judge Papers Pairwise Preference Papers Tool-Use Evaluation Search

Abstract

Language models struggle to generalize beyond pretraining context lengths, limiting long-horizon reasoning and retrieval. Continued pretraining on long-context data can help but is expensive due to the quadratic scaling of Attention. We observe that most tokens do not require (Global) Attention over the entire sequence and can rely on local context. Based on this, we propose L2A (Learning To Attend), a layer that enables conditional (token-wise) long-range memory access by deciding when to invoke global attention. We evaluate L2A on Qwen 2.5 and Qwen 3 models, extending their effective context length from 32K to 128K tokens. L2A matches the performance of standard long-context training to within 3% while skipping Global Attention for $\sim$80% of tokens, outperforming prior baselines. We also design custom Triton kernels to efficiently implement this token-wise conditional Attention on GPUs, achieving up to $\sim$2x improvements in training throughput and time-to-first-token over FlashAttention. Moreover, L2A enables post-training pruning of highly sparse Global Attention layers, reducing KV cache memory by up to 50% with negligible performance loss.

What we could verify

These are the protocol signals we could actually recover from the available paper metadata. Use them to decide whether this paper is worth deeper reading.

Human Feedback Types

missing

None explicit

No explicit feedback protocol extracted.

"Language models struggle to generalize beyond pretraining context lengths, limiting long-horizon reasoning and retrieval."

Evaluation Modes

partial

Automatic Metrics

Includes extracted eval setup.

"Language models struggle to generalize beyond pretraining context lengths, limiting long-horizon reasoning and retrieval."

Quality Controls

missing

Not reported

No explicit QC controls found.

"Language models struggle to generalize beyond pretraining context lengths, limiting long-horizon reasoning and retrieval."

Benchmarks / Datasets

missing

Not extracted

No benchmark anchors detected.

"Language models struggle to generalize beyond pretraining context lengths, limiting long-horizon reasoning and retrieval."

Reported Metrics

partial

Context length

Useful for evaluation criteria comparison.

"Language models struggle to generalize beyond pretraining context lengths, limiting long-horizon reasoning and retrieval."

Benchmarks and datasets

No benchmark or dataset names were extracted from the available abstract.

Reported metrics

context length
Human feedback details
Uses human feedback
No
Feedback types
None
Rater population
Not reported
Expertise required
General
Evaluation details
Evaluation modes
Automatic Metrics
Agentic eval
Long Horizon
Quality controls
Not reported
Evidence quality
Low
Use this page as
Background context only

Research brief

Metadata summary

Language models struggle to generalize beyond pretraining context lengths, limiting long-horizon reasoning and retrieval.

Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.

Key takeaways

  • Language models struggle to generalize beyond pretraining context lengths, limiting long-horizon reasoning and retrieval.
  • Continued pretraining on long-context data can help but is expensive due to the quadratic scaling of Attention.
  • We observe that most tokens do not require (Global) Attention over the entire sequence and can rely on local context.

Researcher actions

  • Compare this paper against nearby papers in the same arXiv category before using it for protocol decisions.
  • Validate inferred eval signals (Long-horizon tasks) against the full paper.
  • Use related-paper links to find stronger protocol-specific references.

Caveats

  • Generated from abstract + metadata only; no PDF parsing.
  • Signals below are heuristic and may miss details reported outside the abstract.

Recommended queries

Contribution summary

  • Based on this, we propose L2A (Learning To Attend), a layer that enables conditional (token-wise) long-range memory access by deciding when to invoke global attention.
  • We evaluate L2A on Qwen 2.5 and Qwen 3 models, extending their effective context length from 32K to 128K tokens.
  • L2A matches the performance of standard long-context training to within 3% while skipping Global Attention for \sim80% of tokens, outperforming prior baselines.

Researcher checklist

  • Human feedback protocol is explicit

    No explicit human feedback protocol detected.

  • Evaluation mode is explicit

    Detected: Automatic Metrics

  • Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Benchmark or dataset anchors are present

    No benchmark/dataset anchor extracted from abstract.

  • Metric reporting is present

    Detected: context length